PHD Thesis-Peter

Colour Measurement and Print Quality Assessment
in a Colour Managed Printing Workflow
Doctoral Dissertation by
Peter Nussbaum
The Norwegian Color Research Laboratory

Faculty of Computer Science and Media Technology
Gjøvik University College
P. O. Box 191
N-2802 Gjøvik, Norway
Submitted to the Faculty of Mathematics and Natural Science at the

University of Oslo in partial fulfillment of the requirements for the
degree of Philosophiae Doctor (PhD) in Imaging Science
Department of Informatics
University of Oslo
P. O. Box 1080 Blindern
N-0316 Oslo, Norway
December 2010
© Peter Nussbaum, 2011
Series of dissertations submitted to the

Faculty of Mathematics and Natural Sciences, University of Oslo
No. 1066
ISSN 1501-7710
All rights reserved. No part of this publication may be

reproduced or transmitted, in any form or by any means, without permission.
Cover: Inger Sandved Anfinsen.

Printed in Norway: AIT Oslo AS.
Produced in co-operation with Unipub.

The thesis is produced by Unipub merely in connection with the
thesis defence. Kindly direct all inquiries regarding the thesis to the copyright
holder or the unit which grants the doctorate.
Abstract
When digital image data are reproduced and printed, the corresponding visual stimulus,
i.e., a print viewed in a certain way, can be highly dynamic. There is a great potential for
variations in the final perceived print quality due to various factors involved in the printing
process, as well as in the viewing of the resulting print. A number of studies and research
have been done in the field of print quality and print assessment, and repeatedly it has been
concluded that it is a very complex issue. Often, printed matter, such as magazines and
newspapers show variations in their print quality caused by a range of factors, and the
competitive situation in the graphic arts and printing industry is an invitation to develop
methods and procedures to limit these variations in the printing process.
The purpose of this study is to develop methodologies and procedures for print
quality assessment in a colour managed printing workflow. Even though multiple factors
can affect the appearance of the print, the aspect "colour" is paid most attention to in this
work. In order to achieve this goal, the work requires collection of relevant data gathered
from visual observations using psychophysical experiments and quantitative measurement
methods, applying state of the art measurement instruments. An important part of the thesis
is the examination of typical measurement devices used in the graphic arts and printing
industry in terms of their accuracy and reproducibility. The performance of measurement
instruments can severely affect the process control and finally the judgement of the print
and proof quality. Hence, the inherent uncertainty of measurement instruments is
investigated and the consequences discussed. Considering the comparison of densitometric
and planimetric measurement technique for newspaper printing the outcome showed that
the proposed model does not accurately describe the relation between the two
measurement technologies due to the large uncertainty for dot meters. This can be
explained by the poor repeatability performance for dot meters on newspaper print.
Regarding the identified measurement variations between different colour
measurement instruments, a solution is required to reduce the measurement uncertainty
between them. Therefore another method is proposed to reduce these variations by
applying a regression technique directly on the measurement output values in the CIELAB
colour space. Simultaneously, the aim is to improve the colorimetric performance and
inter-instrument and inter-model agreement.
V
The main contribution of this thesis is to identify methodologies and routines to
implement standardization procedures in a colour managed printing workflow to ensure
consistent and predictable print quality. In particular, important parts of the result is the
demonstration of a method for soft proofing in a standardized printing workflow and
pointing out procedures for applied colour management in a standardized heat-set web
offset and newspaper printing. The analysis of measurement uncertainty and the proposed
solution to correct the instrument’s systematic errors is also one of the important
achievements in this work.
In conclusion the results can be used as an application independent framework for
those involved in the process of print quality assessment.
VI
The Publications
The seven papers listed below constitute the core work of the present thesis. Further, four
publications and their contribution are related to the present research work, but are not
included in the full text in this thesis.
List of Included Papers
Paper A
P. Nussbaum, J. Y. Hardeberg, and S. E. Skarsbø
Print Quality Evaluation for Governmental Purchase Decisions
In Advances in Printing Science and Technology: Proceedings of the 31st International
iarigai Research Conference, Volume 31, pp. 189-200. M. Lovreček, Ed., Acta Graphica
Publishers, 2004.
Paper B
P. Nussbaum and J. Y. Hardeberg
Print Quality Evaluation and Applied Colour Management in Heat-set Web Offset
In Advances in Printing and Media Technology: Proceedings of the 33rd International
Research Conference of iarigai, Volume 33, pp. 399-411. N. Enlund and M. Lovreček,
Ed., Acta Graphica Publishers, 2006.
Paper C
A. Sole, P. Nussbaum, and J. Y. Hardeberg
Implementing ISO 12646 Standards for soft Proofing in a Standardized Printing
Workflow according to PSO
In Advances in Printing and Media Technology: Proceedings of the 37th International
Research Conference of iarigai, Volume 37, pp. 215-226. N. Enlund and M. Lovreček,
Ed., International Association of Research Organizations for the Information, Media and
Graphic Arts Industries, 2010.
IX
Paper D
P. Nussbaum and J. Y. Hardeberg
Print Quality Evaluation and Applied Colour Management in Coldset Offset
Newspaper Print
Color Research & Application, Wiley.
Article first published online: March 8th 2011, DOI: 10.1002/col.20674
Paper E
M.S. Wroldsen, P. Nussbaum, and J. Y. Hardeberg
A Comparison of Densitometric and Planimetric Measurement Techniques for
Newspaper Printing
In TAGA Journal of Graphic Technology, Technical Association of the Graphic Arts,
Vol. 4, pp. 101-115, 2008.
Paper F
P. Nussbaum, A. Sole, and J. Y. Hardeberg
Analysis of Colour Measurement Uncertainty in a Colour Managed Printing
Workflow
Accepted for publication in Journal of Print and Media Technology Research, The
International Association of Research Organizations for the Information, Media and
Graphic Arts Industries (iarigai).
Paper G
P. Nussbaum, J. Y. Hardeberg, and F. Albregtsen
Regression based Characterization of Colour Measurement Instruments in Printing
Applications
In Electronic Imaging: Color Imaging XVI: Displaying, Processing, Hardcopy, and
Applications, SPIE Proceedings, 7866, San Francisco, CA, 2011.
X
List of Related Papers
• P. Nussbaum, A. Sole, and J. Y. Hardeberg

Consequences of using a number of different color measurement instruments in a
color managed printing workflow
In Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 61st
Annual Meeting, New Orleans, LA, 2009.
• P. Nussbaum and J. Y. Hardeberg

Print quality evaluation and applied colour management in coldset offset newspaper
print
In Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 60th
Annual Meeting, San Francisco, CA, 2008.
• S. Roch, J. Y. Hardeberg, and P. Nussbaum

Effect of time spacing on the perceived color
In Proc. SPIE: Color Imaging XII: Processing, Hardcopy, and Applications, 2007,
6493, pp. 185-195.
• J. Y. Hardeberg, P. Nussbaum, S. Roch, and O. Panak

Time matters in soft proofing
In ACTA GRAPHICA, Journal for Printing Science and Graphic Communication, vol.
19, pp. 1-10, 2007.
XI
Contents
Abstract ......................................................................................................... V
Acknowledgements....................................................................................VII
The Publications ......................................................................................... IX

List of included Papers ................................................................................................... IX
List of related Papers ...................................................................................................... XI
PART I INTRODUCTION
1 Introduction .............................................................................................3
1.1 Motivation ................................................................................................3

1.2 Aim of the Study ......................................................................................3
1.3 Research Methodology ............................................................................4
1.4 Outline of Thesis ......................................................................................5
2 Background ..............................................................................................7
2.1 Colour Fundamentals..............................................................................7

2.1.1 Human Visual System ..........................................................................................7
2.1.2 Colorimetry ..........................................................................................................8
2.1.3 CIE Standard Observer.........................................................................................9
2.1.4 CIE Standard Illuminants ...................................................................................11
2.1.5 Metamerism........................................................................................................12
2.1.6 CIE 1931 XYZ Colour Space.............................................................................13
2.1.7 1976 Uniform Colour Spaces .............................................................................16
2.1.8 CIELAB..............................................................................................................17
2.1.9 CIELUV .............................................................................................................19
2.1.10 CIELAB Colour Difference ...............................................................................20
2.2 Colour Measurement.............................................................................27

2.2.1 Measurement Instruments and their Application ...............................................27
2.2.2 Sources of Error and Measurement Uncertainty ................................................37
XIII
2.3 Image Reproduction and Colour Management ................................. 41
2.3.1 Additive or Subtractive Colour Mixing .............................................................41
2.3.2 Image Reproduction ...........................................................................................42
2.3.3 Colour Gamut Mapping .....................................................................................45
2.3.4 Principles of ICC Colour Management..............................................................46
2.3.5 Calibration and Characterization........................................................................49
2.3.6 Standardization in Offset Printing and Prepress ................................................50
2.4 Print Quality and Print Assessment.................................................... 55

2.4.1 What is (Print) Quality? .....................................................................................56
2.4.2 Factors Contributing to Print Quality.................................................................58
2.4.3 Print Quality Evaluation by Visual Observation................................................60
2.4.4 Quantitative Print Quality Evaluation ................................................................70
3 Summary of included Papers ............................................................... 75
3.1 Paper A: Print Quality Evaluation for Governmental Purchase

Decisions ................................................................................ 77
3.2 Paper B: Print Quality Evaluation and Applied Colour Management in

Heat-set Web Offset............................................................... 81
3.3 Paper C: Implementing ISO 12646 Standards for soft Proofing in

a Standardized Printing Workflow according to PSO............ 85
3.4 Paper D: Print Quality Evaluation and Applied Colour Management

in Coldset Offset Newspaper Print ........................................ 89
3.5 Paper E: A Comparison of Densitometric and Planimetric

Measurement Techniques for Newspaper Printing .............. 93
3.6 Paper F: Analysis of Colour Measurement Uncertainty in a Colour

Managed Printing Workflow ................................................. 97
3.7 Paper G: Regression based Characterization of Colour

Measurement Instruments in Printing Applications ............ 101
XIV
4 Discussion and Conclusion..................................................................105
4.1 Discussion of Papers in Context .........................................................105

4.2 Summary of Contributions .................................................................109
4.3 Conclusion ...........................................................................................113
4.4 Perspectives ..........................................................................................115
References ..................................................................................................119
PART II INCLUDED PAPERS
Paper A………………………………………………………………………..…139
Paper B………………………………………………………………………..…161
Paper C………………………………………………………………………..…183
Paper D………………………………………………………………………..…207
Paper E………………………………………………………………………..…231
Paper F………………………………………………………………………..…253
Paper G……………………………………….………………………………….289
XV
XVI
PART I
INTRODUCTION
1
2
Introduction
1 Introduction
This chapter provides a brief introduction considering the aim of and the methodology
used in the present work, and also describes the structure of the thesis.
1.1 Motivation
Although the electronic media market is in rapid development, it is still considered that
most people prefer reading on paper rather than on displays [75]. From an environmental
point of view recent research results demonstrate that paper based media are contributing
less to global warming compared to electronic media, taking a life cycle perspective on the
products into account [40, 119, 120]. These, among many other reasons, are why paper
based material still has a notable presence in the communication market, and printed
matter such as books, textbooks, magazines and newspapers still play an important role in
our daily life. A recent study by Intergraf1 [76] considering the future of the European print
industry confirms this statement. However, the study concludes that the effect of external
factors, such as expansion in communication technology, market shifts, and changes in
production technologies, will force traditional printing companies to be more agile and to
respond more swiftly to future changes. The competitive situation has become very
dynamic in the sense that not only the competition with other media has grown but also the
requirements for more rapid, predictable and cost efficient print production has increased.
To meet these demands, standardization and the adoption of international specifications in
the printing workflow and process control are essential and an important key to success.
We recognized the potential in facing the opportunities to embrace these challenges and
decided to approach this research field. Print quality, print assessment, process control,
colour management, and colour measurement are the main issues involved and discussed
in this work.
1.2 Aim of the Study

The aim of this study is to develop methodologies and procedures for print quality
assessment in a colour managed printing workflow according to objective evaluation and
perceptual judgement. Furthermore, due to the quantitative evaluation involved, an
1
International confederation for printing and allied industries (Intergraf). http://www.intergraf.eu
3
Introduction
important part of the study is the investigation of measurement devices in terms of

accuracy and reproducibility for process control and print and proof quality assessment.
Hence, the measurement uncertainties of measurement instruments used in the graphic arts
and printing industry will be investigated and the consequences discussed. The challenges
identified during the initial research period in using different measurement instruments and
the reliability of the obtained measurement data have led to finding a solution to reduce the
measurement uncertainty among different instruments. Hence, a method is proposed to
reduce the colour differences among different measurement instruments, and
simultaneously improve the colorimetric performance and inter-instrument and inter-model
agreement.
In this study print quality is seen in the context of a colour managed printing
workflow and the factors affecting the appearance. Although there are a number of factors
in the workflow influencing the appearance of the print, it is in particular the factor
“colour” which is the most important focus of attention in this work.
Our aim has from the outset of the project been to develop methodology and
procedures that could be used as an application independent framework for those involved
in the process of print quality assessment.
1.3 Research Methodology

The research area of the thesis is in the field of colour measurement and print evaluation in
a colour managed printing workflow. In order to obtain the stated aim of the study which
includes the development of methodologies and procedures for print quality assessment,
our research methodology requires gathering relevant data from visual observations using
psychophysical experiments and quantitative measurement methods applying state of the
art measurement instruments. It has been of pivotal importance to obtain measurements
from different types of instruments used in the graphic arts and printing industry, and to
study inter-instrument and inter-model performance in terms of accuracy and
reproducibility.
In the research period three different types of psychophysical methods, presenting
the printed samples to the observers and collecting the observer’s judgement were used in
Paper A [134], Paper B [132], and Paper D [131]. On the other hand considering the
instrumental measurements, different types of measurement instruments were used to
gather the data set in all seven Papers A-G [131-134, 136, 172, 194].
4
Introduction
1.4 Outline of Thesis

The thesis is intended to provide the reader with the understanding needed to determine
print quality in a colour managed printing workflow using both psychophysical experiment
and quantitative evaluation. In order to structure the content the thesis is divided into two
parts and four chapters.
PART I provides in Chapter 2 an overview of the research area and related fields including
an introduction to colour fundamentals, colour measurement, colour reproduction
and colour management, and print quality and print assessment which are
essential topics of the thesis. Chapter 3 highlights the research work by
summarizing the contribution from the individual papers. Furthermore, a
discussion, summary, and conclusion of the entire thesis are presented in Chapter
4 and finally suggestions for further research are proposed.
PART II presents the main contribution of the thesis including the seven published papers
describing the research work of the thesis.
5
Introduction
6
Background Colour Fundamentals
2 Background
The focus of the present study is in the field of print quality evaluation in a colour
managed printing workflow. Therefore some knowledge of fundamental colour principles
is important and the subject has to be addressed. Hence, the purpose of this chapter is to
give an overview of the research area and provide a concise introduction to some of the
fundamentals of colour science, colorimetry, colour measurement, colour management and
print quality.
2.1 Colour Fundamentals

In this chapter we recall some of the basic concepts of measuring colours, in particular the
relationship between light sources, reflective objects and observers. For instance we
explain how to compute the tristimulus values CIEXYZ and CIELAB of an object,
knowing its spectral reflectance property, illuminant and observer conditions.
However, a wide range of comprehensive literature exists describing in depth the
different topics addressed in this section. For instance Roberts [154], and Wandell [188]
review the properties of the human visual system, and Valberg [185] explores the
fundamentals of human colour vision. A complete treatise on colour science can be found
in the book by Wyszecki and Stiles [198]. Further, refer for example to Hunt [70], [71] for
measuring colour and for colour reproduction, respectively, and Otha and Robertson [139]
for colorimetry, Schanda [159] for understanding the CIE system, and Sharma [167] for
colour management.
2.1.1 Human Visual System

The human visual system is sensitive to electromagnetic radiation in the range of
wavelengths of approximately 380 nm-780 nm, usually referred to as light. When we
observe light reflected from an object as illustrated in Figure 1, the light enters the eye and
is focused onto the retina, which consists of a mosaic of specialized cells called rods and
cones containing pigments that respond to light. When the visual pigments absorb light
chemical changes take place, which initiate electrical impulses. The stimuli are processed
by a neural network of brain cells and eventually lead to the excitation of other cells in
various specialized areas of the outer region of the brain known as the visual cortex where
the stimulus is interpreted as a sensation of colour [191]. In the human retina there are
7
Colour Fundamentals Background
approximately 130 million photoreceptors (120 x 106 rods and 6 x 106 cones). The rod
cells are responsible for low-intensity night vision referred to as scotopic vision, while at a
higher or photopic level of illumination three types of cone receptor cells have maximum
sensitivity to light in three different parts of the visible spectrum at 420 nm (short
wavelengths corresponding to violet), 530 nm (medium wavelengths corresponding to
yellowish-green) and 560 nm (long wavelengths corresponding to greenish-yellow) named
as S, M, and L cones respectively [185]. It is important to note that the three cone types are
not distributed evenly throughout the retina. The ratio of their relative abundances is
approximately twice as many L cones as there are M cones. The S cones are rare
throughout the retina and are almost entirely absent in the fovea itself [67].
Figure 1: Schematic representation of the principles of viewing colour.
The perceived colour stimulus depends on the spectral power distribution (SPD) of the
light source, the spectral reflectance properties of the object, the spectral sensitivities of the
L, M, and S cones in the human eye and the resulting interpretation of the sensation by the
visual system.
2.1.2 Colorimetry
To quantify and to model the human colour perception, colorimetry has been introduced,
which is the science of measuring colour. This takes into account the physical
characteristics of the light source, the object’s spectral properties and the physiological
aspects of human vision [138]. In colorimetry this is also referred to as psychophysical
interaction, which is the relationship between the physical properties and the resulting
colour sensation. Colorimetry is founded on a classic series of colour matching
experiments that allowed the trichromatic (three-channel) properties of human vision to be
studied and characterized. Hunt [70] gives a very good overview of CIE colorimetry [25],
8
and most of the present summary is based on that source (all formulae are taken from
there, unless stated otherwise).
In 1931 the international Commission on Illumination (Commission Internationale
de l’Eclairage, CIE) introduced a system for the specification of colour measurement based
on additive colour mixing. Thus, by choosing three appropriate primaries all colour stimuli
can be matched by the additive mixture. In conducting a colour matching experiment, the
observer regulates the amount (intensities) of the three primaries until their mixture
appears to match the test colour, and the amount of the primaries used to produce the
match are commonly known as the tristimulus values. A set of three curves called colour-
matching functions (CMF) are obtained by using a series of test colours of monochromatic
light throughout the visible spectrum and recording the amounts of the three primaries
required, to match each individual test colour.
2.1.3 CIE Standard Observer

The 1931 CIE system is based upon a linear transformation of the original colour-matching
functions averaged from matching experiments performed by Guild [62] and Wright [192]
using particular real primaries R, G, B to a set of imaginary primary stimuli known as X, Y
and Z. The CIE experiment for deriving the colour-matching functions were conducted at
2º viewing angle, and the functions are denoted by x ( λ) , y ( λ) , and z ( λ ) . In 1964 CIE
defined Supplementary Standard Colorimetric Observer colour- matching functions for
samples viewed under 10°, denoted by x10 ( λ) , y10 ( λ) , and z10 ( λ) . The viewing angle
refers to the angle formed at the eye by an object at normal viewing distance
(approximately 45 cm) from the viewer. When viewing images either on the monitor or on
print the viewing angle created is considered as 2 degree. Therefore, computing colour
measurement data for print quality evaluation or colour management the CIE 2º (1931)
Standard Observer is required. Figure 2 to the left illustrates the colour-matching functions
for the CIE 1931 Standard Colorimetric Observer, expressed in terms of matching stimuli
R, G, and B, with monochromatic primaries of 435,8 nm, 546,1 nm, and 700 nm
respectively. Since R, G, and B were real primaries some of the values were negative. To
ensure that tristimulus values for all real colours are always positive, the CIE adopted three
unreal primaries X, Y, and Z. The transformed system is called the X, Y, Z system with
colour-matching functions x ( λ), y ( λ ), and z ( λ ) as shown in Figure 2 to the right. This
system is often referred to as the CIE colour-matching functions for the CIE 1931 Standard
Colorimetric Observer or the 2° Observer.
9

b̄(λ) z̄(λ)
r̄(λ)
Spectral tristimulus values

Spectral tristimulus values

ḡ(λ)
ȳ(λ)

x̄(λ)

Wavelength Wavelength
Figure 2: To the left: Colour-matching functions for the CIE 1931 Standard Colorimetric
Observer, expressed in terms of matching stimuli R, G, and B, with monochromatic
primaries of 435,8 nm, 546,1 nm, and 700 nm respectively (Data set available by
Rijgersberg [153]). To the right: CIE colour-matching functions for the CIE 1931
Standard Colorimetric Observer (full lines), and for the 1964 CIE Supplementary
Standard Colorimetric Observer (dashed lines), (Data set available by CIE [31]).
The tristimulus values X, Y, and Z for a given object, which is illuminated by a

certain light source, can be calculated for the CIE Standard Observer by summing the
products of these distributions over the wavelengths range typically from 380 nm to 780
nm, usually at 5 nm intervals. Figure 3 shows a schematic of the computing process.
Light source Object CIE Standard Observer Tristimulus values

X

Tristimulus value
Relative power
Reflectance
Y

X X =
Z

Wavelength Wavelength Wavelength
Figure 3: Data for the light source, the object, and the human observer are used to derive
the tristimulus values CIEXYZ.
These three numbers constitute the units of the first CIE colour space – CIEXYZ, whose
coordinates are referred to as tristimulus values. The calculation of the CIE 1931
tristimulus values X, Y, and Z is shown in the following equations:
10
X = k ∑ x ( λ )R( λ )S( λ )Δλ

Y = k ∑ y ( λ )R( λ )S( λ )Δλ
Z = k ∑ z ( λ)R( λ)S( λ)Δλ (1)
100
k=
∑ y (λ)S(λ)Δλ
where X, Y, and Z are the CIE tristimulus values; x ( λ) , y ( λ) , and z ( λ ) are CIE 1931
Standard Observer colour-matching functions; S(λ) is the spectral power distribution of the
light source; R(λ) is the spectral reflectance of an object and k is a normalising constant.
Δ(λ) is the wavelength interval (e.g. 5, 10 or 20 nm).
Note that by convention, k is determined in such a way that Y=100 when the object
is a perfect diffuse reflector (for which R(λ) =1 for all λ), which implies that the absolute
spectral power distribution for the light source (illuminant) is not required.
Further, note that the X, Y, Z colour matching functions do not correspond to a set
of physical primaries nor represent the actual response properties of the cones, but are
linear transformations of the R, G, B colour matching experiment using real primaries. The
replacement of the R, G, B primaries by the three “imaginary” primaries X, Y, Z was
completed to avoid negative numbers in the colour specification and normalized to yield
equal tristimulus values for the equi-energy spectrum. Furthermore, y ( λ) was chosen to
be equivalent to the luminous efficiency function for photopic vision, V(λ).
2.1.4 CIE Standard Illuminants

As seen above the perceived colour by a human observer is the result of the interaction
between the colour-matching functions, the spectral reflectance characteristic of the object,
and very importantly by the spectral power distribution (SPD) of the light source under
which the sample is viewed. Therefore CIE has specified a range of standard illuminants
(light sources) in terms of their spectral power distribution for industrial applications. In
1931 CIE defined the illuminant A (tungsten light) and in 1963 CIE introduced D50
(representing daylight) with a correlated colour temperature (CCT) of 5000K, and D65
(with a CCT of 6500K) [70]. Most important for the graphic arts and printing industry D50
is the common illuminant required for the appropriate viewing conditions mentioned in
ISO 3664 [79].
11
2.1.5 Metamerism
As discussed previously, the trichromatic properties (three colour receptors) of the human
visual system means that the detection of all colours is reduced to three sensory quantities.
The consequence of that characteristic is the fact that, although two colour stimuli have
different spectral power distribution, the visual appearance can be identical. A visual
colour match can for example be obtained between a colour stimulus on a printed substrate
and a corresponding colour stimulus on a monitor display (e.g. soft proofing), although
their spectral power distributions obviously are very different. This phenomenon, called
metamerism, can be explained by the individual cone responses to the cumulative power
energy across a broad range of wavelengths so that different mixtures of light across all
wavelengths can produce an equivalent stimulation on the receptor and eventually result in
the same tristimulus values or colour sensation. For further information on metamerism,
see e.g. [25, 42, 71, 198].
In other words, when two samples with different spectra appear to match under one
light source but are different under another light source, this phenomenon is known as
metamerism, and the samples are known as a metameric pair. This can lead to unwanted
practical problems, e.g. in the graphic arts industry, where image reproduction issues such
as print and proof quality may be evaluated under inappropriate viewing conditions.
Typically, a visual comparison between a print and a proof under office light condition can
result in an unacceptable mismatch whereas under standard viewing conditions using D50
the comparison results in a visual match. However, even though the phenomenon may
cause some problems sometimes, metamerism makes colour image reproduction
practically possible by using three primary colours to obtain a colorimetric match rather
than recreating the spectral reflectance of the original colour.
In print quality and soft proofing the appropriate light source is required to predict
the appearance. The CIE has published standards for assessing quality of light sources and
has recommended the use of metamerism indices which quantify the degree of
metamerism by changing the conditions [25]. Firstly, the illuminant metamerism index
considers the colour difference between two samples, which match under one illuminant
but will mismatch under another illuminant. CIE recommends a D65 illumination as the
reference illuminant and the test illuminant can be chosen depending on the application.
Typically ISO 3664 [79] specifies the viewing conditions for printed images as well as
images displayed in isolation on colour monitors according to the metamerism index in
accordance with ISO/CIE 23603 [95].
12
Furthermore, there are three other types of metamerism to consider. The observer
metamerism index quantifies the colour difference degree in which the CIE standard
observers and standard deviate observer do not agree in matching. Field-size metamerism
is to consider in order quantifying the degree of colour difference of a sample pair by
changing the viewing field size. For instance, two samples can be deemed identical in 2
degree viewing condition, but different under 10 degree viewing field. Finally, when the
colour matching disappears by changing the illumination and viewing conditions the
geometrical metamerism has to be considered, which usually is seen in metallic paint
where two samples are identical in a given viewing geometry while different in other
conditions [70].
2.1.6 CIE 1931 XYZ Colour Space

As seen above, the CIE X, Y, Z tristimulus values are fundamental measures of colour.
Because they do not express immediately an obvious representation of colours nor do they
have a direct perceptual correlation, the X, Y, Z tristimulus values can be transformed into a
graphical system known as the CIE x, y chromaticity diagram, providing a two-
dimensional representation. In this system a colour is represented by its coordinates x, y,
known as chromaticity coordinates, derived from the tristimulus values as follows:
X
x=
X +Y + Z
Y
y= (2)
X +Y + Z
Z
z=
X +Y + Z
The derived x and y values are normalized tristimulus values such that x + y + z = 1. The
variables x and y are projective coordinates forming the chromaticity diagram, representing
a visualization of the three-dimensional XYZ colour space onto a two-dimensional plane.
The chromaticity of all colours can be plotted in the chromaticity diagram. But stimuli of
identical chromaticity but different luminance are collapsed onto the same point in the two-
dimensional plane of the chromaticity diagram. Consequently, the luminance value must
be quoted as a separate value (Y). The CIE xyY colour space is widely used to specify
colours in practical application. In photometry, the tristimulus value Y has units either of
lux (lx) which is lumens per square meter, or candelas per square meter, cd/m2. The unit
lux is in particular used to describe the concentration of luminous flux in a given direction,
at a given point of a real surface (e.g. 2 000 lx in pressroom for critical comparison, P1)
13
[79]. The unit cd/m2 is intended to express how much light a device can generate (e.g.
monitor display) [83].
Spectral locus
546.1 nm
700.0 nm
Purple boundary
435.8 nm
Figure 4: CIE x, y chromaticity diagram including the chromaticity coordinates of the real
RGB primaries (forming triangle of dashed lines) corresponding to the colour-matching
functions of r ( λ ),g ( λ), and b ( λ) .
Figure 4 illustrates the CIE x, y chromaticity diagram with the horseshoe-shaped outline,
representing the chromaticity of monochromatic light called the spectral locus; with
wavelengths shown in nanometers along the curve. Colours, which are less saturated
appear in the interior of the diagram with white at the centre. The corners of the triangle
(dashed lines) shows the location of the chromaticity coordinates of the real RGB primaries
corresponding to the colour-matching functions of Figure 2. An important characteristic of
the CIE x, y chromaticity diagram is the additive colour mixture calculation.
If any two points on the chromaticity diagram are chosen, then all colours which lie
on a straight line between the points, can be obtained by mixing these two colours. The
straight line which connects the two end points of the spectrum locus is known as the
purple boundary and presents the additive colour mixing of the monochromatic lights red
and blue (violet). The diagram represents the gamut of all the chromaticities visible to the
average person and all colours can be formed by mixing three sources and are found inside
the area enclosed by the spectral locus and the purple boundary.
14
Although the CIE x, y chromaticity diagram is often used in colour management,

e.g. to represent certain primary colours (e.g. chromaticity coordinates of sRGB) and to
visualize the corresponding gamut, it suffers from a serious disadvantage. In fact, the
distribution of the colours in the chromaticity diagram is not uniform. The consequence of
this is that colour differences, represented by equal distances in the CIE x, y chromaticity
diagram, are not perceived as being equal, as shown in Wright’s and MacAdam’s results.
In Wright’s experiment performed 1941, he shows that each line in the CIE x, y diagram
represents a colour difference of equal proportion. However, the length of lines varies
depending on the region of the diagram as seen in Figure 5 a). In a perceptual uniform
colour space they are expected to have the same length. In 1942 MacAdam conducted
visual experiments to prove the non-uniformity of the CIE x, y chromaticity diagram and to
determine the perceptual distance within the CIE x, y chromaticity diagram [113]. He
performed additive colour mixing experiments for 25 different test colours, establishing
elliptical regions on the chromaticity diagram which contain all colours that are perceptual
indistinguishable from the colour at the centre of the ellipse. According to his results, the
size and shape of these regions vary over the chromaticity diagram. Although there are
very small perceptible differences in the blue-purple colours region, in the green area of
the CIE x, y chromaticity diagram, where colour differences are not distinguishable, the
regions are approximately 10 times larger. Figure 5 b) demonstrates the different ellipses,
which are called the MacAdam ellipses. Ideally, the MacAdam ellipses should all be
circles of the same radius if the chromaticity diagram was perceptually uniform.
To overcome these shortcomings a number of different diagrams have been
proposed by squeezing or stretching the CIE x, y chromaticity diagram to provide a better
correlation between the perceived colour difference and the corresponding distances in the
diagram.
15
a.) b.)
Figure 5: a) CIE x, y chromaticity diagram with lines representing small colour
differences. b) MacAdam (1942) ellipses plotted in CIE x, y chromaticity diagram (axes of
the plotted ellipses are 10 times their actual lengths).
2.1.7 1976 Uniform Colour Spaces

To correct the non-uniformity in the x, y chromaticity diagram, in 1976 the CIE
standardized the u’, v’ diagram as an approximately uniform chromaticity diagram
whereby its axes are defined as follows:
4x
u' =
−2x +12y + 3
(3)
9y
v' =
−2x +12y + 3
Figure 6 demonstrates the derived CIE standardized uniform CIE u’, v’ chromaticity
diagram in which a selection of lines seen in Figure 5 a) are shown again. It can be
immediately observed, that the variation in the length of the lines has been much reduced.
In fact, the ratio of the longest line to the shortest line is only about four to one, compared
to the CIE x, y chromaticity diagram where the ratio is approximately twenty to one.
16
Figure 6: CIE standardized uniform CIE u’, v’ chromaticity diagram.
Although the improvements have increased, the perceptual uniformity of the CIE u’, v’
chromaticity diagram it is still only applicable to colours having the same luminance. In
general, however, colours vary in both the chromaticity and luminance domain. Therefore
CIE has introduced a further model combining these variables in another colour space.
2.1.8 CIELAB
In 1976 a major development was made by the CIE introducing the CIELAB uniform
colour space and colour specification intended to be used for surface colours.
The CIELAB colour space is established as a colour-opponent three-dimensional
space with the axes L*, a*, and b* forming a rectangular or Cartesian coordinate space. L*
shows the dimension of lightness, while a*, b* refer to the colour-opponent dimensions
red-green and yellow-blue, respectively [71]. The scale of lightness is from 0 to 100, with
L* of 0 representing black and L* of 100 representing white as demonstrated in Figure 7.
The L*a*b* coordinates are calculated using a non-linear transformation from the
XYZ tristimulus values with the intention to provide a partial solution to both problems of
colour appearance and colour difference in the new space. Hence, the aim of the non-linear
transform of the tristimulus values in the CIELAB formula is to allow the Euclidean
distance between two points to enhance the prediction of the visual colour difference
between the colour stimuli represented by these two points [191].
17
L* = 100 = White
E*ab
+b* = Yellowish
-a* = Greenish +a* = Reddish
-b* = Blueish L* = 0 = Black
Figure 7: Schematic diagram of CIELAB axis L*, a*, and b* (after Berns, page 71 [16]).
From the obtained tristimulus values X, Y, and Z the L*a*b* coordinates can be calculated
based on the equations:
L*= 116(Y / Yn)1/3 - 16, for Y / Yn > 0.008856,
L*= 903.3(Y / Yn), for Y / Yn ≤ 0.008856,
(4)
a*=500[f(X / Xn) 1/3 - f(Y / Yn)1/3],
b*=200[f(Y / Yn) 1/3 - f(Z / Zn)1/3],
where
f(I)= (I)1/3, if I > 0.008856,

f(I)=7.787(I) + 16 / 116, if I ≤ 0.008856,
where Xn, Yn, and Zn refer to the tristimulus values of the reference white point. Although
the reference white can be specified according to the white point of the medium on which a
colour is presented (relative colorimetry), for surface colours the values of Xn, Yn, and Zn
are usually calculated in accordance with the perfect reflecting diffuser and are therefore
equivalent to the light source itself (absolute colorimetry).
A further variant of the CIELAB colour representation is to calculate the polar
coordinates of the perceptual attributes chroma C*ab and hue-angle hab:
*
Cab = a*2 + b*2 (5)
18
⎛ b* ⎞
hab = arctan⎜ * ⎟ (6)
⎝a ⎠
The CIE 1976 chroma C*ab predicts the chromatic content of a colour increasing from 0
(achromatic colour in the centre), in other words the distance from the L* axis as shown in
Figure 8. Whereas the CIE 1976 hue–angle hab describes a colour’s hue, which ranges
from 0° to 360° following the colours of the spectrum.
+b*
C*ab
h ab
+a*
-a*
-b*
Figure 8: CIELAB a*, b* plane including chroma C*ab as the length of the vector, and the
angle as the hue angle hab.
2.1.9 CIELUV
A further colour space was defined by the CIE at the same time as CIELAB and for the
same purpose of uniformity, the CIE 1976 (L*u*v*) uniform colour space. CIELUV colour
space is currently used for applications such as lighting, computer graphics, and other
applications involving additive colour mixing. Its coordinates are defined by the
transformation:
L*= 116(Y / Yn)1/3 - 16, for Y / Yn > 0.008856,
L*= 903.3(Y / Yn), for Y / Yn ≤ 0.008856,
(7)
u*=13L*(u’- u’n),
v*=13L*(v’- v’n),
where u' and v' are the coordinates of the CIE standardized uniform CIE u’, v’ chromaticity
diagram, and the index n denotes the coordinates of the reference white.
In the last couple of decades CIELAB has become almost exclusively used for
colour specification especially for the graphic arts and printing industry. Furthermore,
CIELAB is the source for the vast majority of work on the prediction of colour difference
and also colour appearance. However, it is important to note that both the CIELAB and
19
CIELUV colour spaces are only approximately perceptually uniform and neither of them is
generally considered as superior to the other. Moreover, there are still significant
differences in the correspondence between the perceived colour difference and the
Euclidean distance between two points, especially in the blue region of the CIELAB space
[169]. Nonetheless, compared to the CIE XYZ tristimulus colour space, both of them offer
significant improvements for print quality evaluation.
2.1.10 CIELAB Colour Difference

In the context of print quality or process control it is required to measure and compare
colours. The colour difference between two colour samples can be quantified by plotting
their coordinates in the three-dimensional CIELAB colour space. The Euclidean distance
of two points is defined by ΔE*ab as seen in Figure 7 and defines how well two samples
match or how much they differ. Hence, two main points can be considered which a colour
difference equation is aiming for. Firstly, a colour difference of unit one ΔE (ΔE=1)
represents approximately a just noticeable difference (JND) for samples seen side by side
under defined viewing conditions, i.e. the smallest colour difference which is possible to
distinguish by the observer. Secondly, the magnitude of difference between the numerical
values representing two colour samples should be proportional to the perceived colour
difference between them [121] (see Figure 9).
∆E=1 ∆E=1 ∆E=1 ∆E=1
= ∆E=2 = ∆E=2 =
= ∆E=4 =
Figure 9: Schematic illustration expressing colour differences (after Morovic’s Figure on

page 27 [121]).
20
The equation describing the Euclidean distance between two sample points in the CIELAB
colour space is,
ΔE*ab = [(ΔL*)2 + (Δa*)2 + (Δb*)2]1/2, (8)
where
ΔL* = L*sample - L* target

ΔL* > 0 means sample is lighter than target
ΔL* < 0 means sample is darker than target
Δa* = a*sample - a* target

Δa* > 0 means sample is redder than target
Δa* < 0 means sample is greener than target
Δb* = b*sample - b* target

Δb* > 0 means sample is yellower than target
Δb* < 0 means sample is bluer than target
Considering industrial applications of colour science it is common that one of the samples
is a target and the other one is a sample that is supposed to be a visual match to the
standard, as discussed in the publications in PART II of the present thesis.
An alternative formulation of the above CIE 1976 colour difference is the
computation of the polar coordinates, writing an equivalent formula in terms of lightness
difference, ΔL*, chroma difference, ΔC*ab, and hue difference, ΔH*ab,
ΔE*ab = [(ΔL*)2 + (ΔC*)2 + (ΔH*)2]1/2, (9)
where
ΔC* = C*sample - C*target
and ΔH*ab is know as the CIE 1976 a, b hue-difference, expressed as,
ΔH*ab = [(ΔE*)2 - (ΔL*)2 - (ΔC*)2]1/2, (10)
The total colour difference, ΔE*ab can be divided up into parts of ΔL*ab, ΔH*ab, and ΔC*ab.
Essentially these components correlate with the perceptual attributes, lightness, hue, and
chroma whose squares sum to the square of ΔE*ab. Note that although the simple
difference between two hue angles can be calculated, Δh*ab does not have the property of
being a perceptual attribute. ΔH*ab is computed as what is “left over” once the lightness
and chroma differences are subtracted from the total colour difference.
21
These simple equations have been and are still used effectively to quantify colour
difference in a wide range of applications, in particular the graphic arts and printing
industry where process tolerances in international standards are still mostly defined as CIE
1976 a, b colour differences [85-87].
It is important to note that these measures only correlate approximately with the
corresponding attributes. Furthermore, the perceived colour is always depending on
viewing conditions, including illumination intensity, the colour of the illumination, and the
characteristics of the surroundings.
However, how do we interpret or classify the ΔE*ab scale? When considering color
differences two types of visual assessments are most prevalent – perceptibility and
acceptability [16]. First, perceptibility thresholds indicate what magnitude of colour
difference is a just noticeable difference (JND). Colour difference equations are set such
that their units correspond to JND’s – hence it is commonly stated that any colour
difference below 1 unit is predicted as not being perceptible for samples viewed side by
side [16, 38, 71, 101]. However, in a study by Mahy et al. [114] the authors found a JND
of ΔE*ab =2.3. A rule of thumb for practical classification of a perceptibility threshold
which corresponds with the ΔE*ab scale, when two colour samples are viewed side by side
is proposed by Schläpfer [161] where the author classifies a ΔE*ab < 0.2 as "Not visible",
ΔE*ab between 0.2 and 1.0 as "Very small", ΔE*ab between 1.0 and 3.0 as "Small", ΔE*ab
between 3.0 and 6.0 as "Medium" and ΔE*ab > 6.0 as "Large".
Second, the acceptability threshold can be seen as a less concrete concept and one
that depends strongly on application and industry. For example in the work of Hardeberg
[64] a rule of thumb for practical interpretation of a ΔE*ab is proposed where errors of
ΔE*ab <3 are classified as “Hardly perceptible”, 3 ΔE*ab < 6 is defined as “perceptual, but
acceptable” and ΔE*ab > 6 as “Not acceptable”. Sharma [167] states that ΔE*ab between 4
and 8 is generally deemed acceptable in prepress and colour imaging. In another study by
Stokes et al. [179] an acceptability threshold of approximately ΔE*ab = 6 was found for
their experimental images and observers. We note the disagreement between these
classifications because the evaluation of quality and acceptability is highly subjective and
depends greatly on the experiences and expectations of observers as well as the application
for which the colour stimuli are intended. A further quality metric, which could be more
suitable for the printer market, would be a tolerance threshold and this issue could be the
subject of future research.
22
As mentioned previously, the perceived magnitude of the colour difference does

not entirely agree with the Euclidean distance in CIELAB colour space. Therefore several
attempts have been made to improve the inconsistency between the perceived colour
difference and the Euclidean distance and hence several alternative colour difference
equations have been proposed. Because CIELAB has become a very common and widely
used colour space most of the alternative equations are based on that colour system.
However, the calculation of the distance between two colours is not the Euclidean distance
itself. To correct the non-uniformity in parts of the CIELAB colour space the colour
difference equations firstly decompose the Euclidean distance into components
corresponding to the differences in the colour attributes of lightness ΔL*, hue, ΔH*, and
chroma ΔC* respectively. Essentially, the colour attributes are weighted differently to
define a new colour difference which corresponds better to the perceived colour difference.
In 1984, Clark et al. [32] (members of the Colour Measurement Committee of the society
of Dyers and Colourists) published one of the first advanced difference formula known as
CMC(l:c) providing among other parameters scaling functions for lightness, L*, and
chroma, C*. This equation has been extensively used but it was never adopted as a CIE
standard [191]. The complexity of the CMC formula has been criticised and after
analyzing a large set of psychophysical data it was suggested that simple weighting
functions, SL, SC, and SH, would be sufficient to improve the perceived colour difference.
Hence, CIE proposed a new formula, which is known as CIE94 [26], and the colour
difference ΔE*94 is calculated by
⎡⎛ ΔLab * ⎞ 2 ⎛ ΔCab * ⎞ 2 ⎛ ΔHab * ⎞ 2 ⎤1/ 2

ΔE * 94 = ⎢⎜ ⎟ +⎜ ⎟ +⎜ ⎟ ⎥ (11)
⎢⎣⎝ kLSL ⎠ ⎝ kCSC ⎠ ⎝ kHSH ⎠ ⎥⎦
where
kL = kC = kH = 1 for reference conditions
SL = 1
SC = 1 + 0.045C*ab,Standard
SH = 1 + 0.015C*ab,Standard
The variables kL, kC, and kH are parametric factors to correct for experimental viewing and
illumination conditions. Except for the textile industry, CIE94(2:1:1), a value of one is
recommended for all parametric factors [70].
23
The weighting functions SC and SH are calculated using the chroma value of the
standard sample specimen C*ab, Standard. When neither sample can be deemed as standard,
the geometric mean chroma of the two samples should be used [191]. Traditionally ΔE has
been measured as a spherical distance. A more realistic measure is based on an elliptical
volume to improve the correlation between the visual and numerical colour difference.
Therefore, the goal of the weighting functions SC and SH in ΔE*94 is to scale down hue and
chroma differences for higher chroma colours in comparison to the traditional ΔE*ab and
thus correct for the predominant deficiency in CIELAB. For a neutral colour along the L*
axis the scaling factor is 1, which means that the colour differences ΔE*94 are identical to
the ΔE*ab Euclidean colour difference. On the other hand, the weighting functions SC and
SH are greater than unity for chromatic colours. Hence, the results of ΔE*94 become smaller
than the results of ΔE*ab Euclidean colour difference [169].
Additional attempts have been made to adjust the colour difference equation to
further improve the uniformity, and in 2001 a new colour difference formula, CIEDE2000
(ΔE*00), was proposed [111] and adopted by the CIE [29]. In addition to the weighting
functions for lightness, chroma, and hue, as already used in ΔE*94, the ΔE*00 colour
difference equation includes a number of further parameters to compensate for the non-
uniformity of the CIELAB colour space. Among the new parameters, a term for improving
the performance in the blue colours and a rescaling factor of the CIELAB a* axis
improving the performance for colours close to the L* axis, can be mentioned. For further
details and the equations the reader is referred to Luo et al. [111]. It is important to notice
that even though small noticeable improvements compared to ΔE*94 could be made in
some parts of the CIELAB colour space the complexity of the ΔE*00 formula may reduce
the performance of the colour difference prediction in other parts of the colour space.
Additional concerns with regard to the CIEDE2000 colour difference formula are given by
Kuehni and Luo [106, 112]. In a later study by Granger, the author even claimed
systematic errors in the CIEDE2000 formula [57]. Nevertheless, in a recent study by Pant
and Farup, the authors conclude that, though the CIEDE2000 significantly improves the
visual colour differences, orientation problems have been observed in the ellipses in the
blue and red region [142]. Therefore, further research is required to improve the rotation
term in ΔE*00 or to solve the concerns with the colour difference equations in general. In
general, regarding the concerns mentioned with colour difference equations, there are
several fundamental questions associated with the CIELAB colour space itself [107].
24
Although the CIE system of colorimetry has been used successfully for almost 80
years it is important to note that the current model is limited to the comparison of samples
which are identical in spatial respect and viewed under equal viewing conditions including
chromatic adaptation, light adaptation, luminance level, background colour, and surround
colour [42]. However, with these restrictions in mind, the model serves sufficiently for the
upcoming discussions and definitions in the present thesis.
25
26
Background Colour Measurement
2.2 Colour Measurement

As we have seen previously, there are three factors needed for colour perception (unless
we have coloured light, then the three factors are reduced to two, the light and the
observer, respectively). Firstly, as illustrated in Figure 1, a light source is illuminating the
object and consequently an observer detects the reflected light from the object and converts
the signals into a response, which the human brain interprets as a colour. In sensory
substitution of the human sensor by an instrumental colour measurement device, the goal
of the instrumental colour measurement is to estimate what an observer sees, the colour
appearance of a stimulus [16].
In the graphic arts and printing industry colour measurement instruments are used
primarily for determining the colour characteristics of objects, such as prints and proofs,
and for measuring the behaviour of colour reproduction input and output devices. In the
context of colour reproduction device calibration and characterization colour measurement
is a very important task determining the appropriate parameters and verifying the
behaviour of the devices. In colour printing process control the goal of obtaining colour
measurements is to verify the specified target or reference values to ensure the
repeatability of the process and the ink uniformity across the printing substrate.
Colour measurement instruments include both filter-based instruments that directly
measure colorimetry and devices, which are designed to measure spectral reflectance or
transmittance such as spectrophotometers and instruments intended to measure spectral
radiance or irradiance such as spectroradiometers.
In the next section we briefly address the most important measurement instruments
used in the graphic arts and printing industry and that are also applied in this work. For
more comprehensive details on colour measurement devices the reader is referred to
literature by Berns [16], Battle [14], Hunt [70], Sharma [169], Schanda [159], Völz [187],
Wyble [195] and CIE publication 130 [28].
2.2.1 Measurement Instruments and their Application

In this sub-section different types of (colour) measurement instruments are presented with
respect to their viewing and illumination properties defined by CIE. For completeness,
other types of measurement instruments used in the graphic arts and printing industry are
also presented and discussed.
27
Colour Measurement Background
Tristimulus colorimeter
The simplest device to measure colours is a tristimulus colorimeter whose spectral
responsivity mimics the colour matching functions of the Standard Observer [160]. It
measures, as suggested by its name, colour tristimuli and reports these as colour values in
CIEXYZ, and CIELAB. The key elements include a light source (for measuring materials)
usually illuminant C or D65, a set of colour filters and commonly a silicon photodiode as a
photo-detector. The ”spectral” separation is obtained by either using colour filters placed in
front of the photodiode or by using spectrally different light sources to illuminate the
sample, often using 45°:0° geometry, i.e. circumferential 45° illumination and normal
viewing [25]. To be able to report CIE colorimetry the sensitivities of the colour filters are
intended to have a close match to a set of colour matching function. As seen in Figure 10,
for filter based colorimeters, four independent filters are usually used, two approximate the
y ( λ) and z (λ ) colour matching functions and two approximate the two humps of the
x ( λ) colour matching functions [169].
Display
CIEXYZ
Standard Observer response

λ
λ
Photodiodes
Colour Filter
λ
Light 0 Light
source source
45
Sample
mple
e
Figure 10: Basic features of a tristimulus colorimeter.
In general, the tristimulus colorimeter is easy to use, it is fast and inexpensive, but
does not provide any detailed spectral information. Nevertheless, colorimeters are often
used in the graphic arts industry to calibrate and characterize self-luminous sources, such
as monitor displays, although there might be some performance restrictions in terms of
accuracy due to the aging of the colour filters and poor reproducibility to agree with the
28
CIE colour matching functions [8, 14]. A study by Gardner compares different calibration
methods for tristimulus colorimeters to reduce the measurement uncertainties [51]. It is
also important to note that for surface colours, the colorimetric values measured are valid
for one illuminant only, the instrument light source, but it is often desirable to know the
tristimulus values or colour difference for different illuminants. Thus a colorimeter cannot
give any indication of metamerism [70].
Spectrophotometer
Nowadays, the most common colour measurement instrument in the graphic arts and
printing industry is the spectrophotometer which measures the ratio of reflected to incident
light (the reflectance) from a sample at many points across the visible spectrum, that is, as
seen in Section 2.1.1, between about 380 nm and 780 nm.
Reflectance = Reflected light / Incident light
The main components of all spectrophotometers for colour measurement include a light
source (for measuring materials), a wavelength selection device, and a photo detector as
shown in Figure 11.
Reflectance
Factor R
Display
Detector Array
Light
source Lens
0
45 Diffraction
Grating
Aperture
Sample
mple
e
Figure 11: Main measurement components of a 45°x:0°geometry spectrophotometer.
Because we are measuring the ratio of incident to reflected light with a spectrophotometer,
the type of light source illuminating the sample should not matter. However, as certain
paper materials contain a fluorescent whitening agent (FWA), they absorb energy at one
part of the spectrum and re-emit at another. Thus the spectral power distribution (SPD) of
the light source can have a dramatic effect on the measured colour. A paper published by
29
Andersson and Norberg [3] points out how important the UV-content of the illuminant can
be when carrying out colour measurements on prints containing FWA; even small changes
in UV-content can affect the measured reflectance spectra. Pulsed xenon, quartz halogen
and tungsten gas filled, type A lamps are common light sources for spectrophotometers
[16].
The reflected or transmitted light is passed on to the wavelength selection device or
spectral analyser, where the light is split into its spectral components to be measured. In a
spectrophotometer, prisms, gratings, and interference filters are the technologies used to
separate light into narrow bands. Prisms and gratings both separate the wavelength
spatially. Although both methods of dispersing light can still be found, today most
spectrophotometers use diffraction gratings. Light passing through the grating will be
diffracted at a fixed angle, which is dependent on its wavelength and the dispersed light is
focused directly onto the photo detector array. Most spectrophotometers work in the range
380 nm - 780 nm with sampling intervals reporting reflectance at 5-, 10-, or 20-nm
intervals.
CIE geometries for illumination and viewing for reflection

The illumination and viewing geometry of a colour measurement instrument is very
important and can affect the way a colour is measured. The CIE has defined some standard
terms and geometrical conditions denoted by symbols, such as d:8° or 45°:0°. The symbol
‘d’ indicates the use of an integrating sphere, and the numbers before and after the colon
indicate the angles of incidence and viewing, respectively. For instruments using an
integrating sphere a glossy sample can be measured in two conditions either with the
specular component included or excluded. Hence, the geometry is distinguished by adding
the letter ‘e’ for excluded or ‘i’ for included. The current CIE [25] recommendations for
reflection measurements with diffuse geometries:
Diffuse Geometries:
• Diffuse: eight-degree geometry, specular component included (di:8°)
• Diffuse: eight-degree geometry, specular component excluded (de:8°)
• Eight degree: diffuse geometry, specular component included (8°:di)
• Eight degree: diffuse geometry, specular component excluded (8°:de)
• Alternative diffuse geometries (d:0°) and (0°:d)
30
• Diffuse/diffuse geometry (d:d)
Directional Geometries
For 45° geometries, the CIE distinguishes between an annular and a directional
arrangement by adding the letters ‘a’ or ‘x’, respectively.
• Forty-five degree directional/normal geometry (45°x:0°)
• Forty-five degree annular/normal geometry (45°a:0°)
• Normal/Forty-five degree directional geometry (0°:45°x)
• Normal/Forty-five degree annular geometry (0°:45°a)
With these CIE recommendations the 45°x:0° geometry replaces typically the 45°/0°
geometry. The 45° geometry spectrophotometer is widely used within the graphic arts and
printing industry, where the sample is illuminated at an angle of 45° and the measurement
takes place at 0° (perpendicular to the sample), see Figure 11. CIE has also recommended
six different standard geometries of illumination and viewing for transmitting material. For
further details on measuring geometries we refer to Schanda [159] and Otha and Robertson
[139].
It is important for an operator to understand the critical parameters of an
instrument. Therefore CIE has worked out a terminology (first published in 2004) that
makes the definition of the measurement geometry easier. CIE has recommended that the
manufactures follow these new guidelines. However, most of the largest manufactures of
colour measurement instruments use still (after six years) the old terminology (e.g.
reflection 45°/0°).
Spectrocolorimeter
ASTM E1347 [7] defines a spectrocolorimeter as a spectrometer which has a dispersive
element (such as a prism, grating, or interference filter) instead of RGB filters that is
normally capable of producing as output only colorimetric data (such as tristimulus values
and derived colour coordinates) but not the underlying spectral data from which
colorimetric data are derived. According to Rich [151] a better characterization accuracy of
the colour and colour difference of metameric pairs can be obtained. This is achieved by
sampling and digitizing the standard illuminant and the measured object at discrete points,
along the visible spectrum, and eventually to calculate the tristimulus values.
31
Spectroradiometer
Radiometry is the science of measurement of radiant energy, including light with respect to
absolute power. Spectroradiometers are designed to measure radiometric quantities
(irradiance2 and radiance3) in a narrow spectral bandpass as a function of wavelength.
They have the same principal components as the spectrophotometer, except for the light
source. They are designed to measure the spectral properties of light sources, such as
viewing booth, monitor displays, and projectors. The tele-spectroradiometer (TSR) is the
most frequently used instrument in this category. The key components of the TSR are a
telescope, a monochromator, and a detector. As for most of the spectrophotometer the
measurement range interval of a TSR is between 380 nm and 780 nm. But the sampling
interval of the TSR often has a spectral resolution ranging from 1 nm to 10 nm, which is
sufficient for colour work using e.g. fluorescent lamps as sources (some light sources have
monochromatic emission lines in their spectra) [169].
The advantage of the TSR is the measurement of the object at a distance
corresponding to its actual observing position including common viewing distance and
viewing conditions. In other words it can measure all forms of colours (surface and self-
luminous), which are particularly important for cross-media reproduction, e.g. to match an
image displayed on a monitor to the output from a printer in the viewing booth (e.g.
verifying typically a soft proofing set up). However, due to the high costs and complexity
the TCR is not a common measurement device used in the graphic arts and printing
industry.
Abridged Spectrometers
According to Berns [16] abridged spectrometers are instruments that sample the visible
spectrum at several discrete bands by e.g. replacing the broadband coloured filters with
narrowband interference filters with approximately 50 nm bandwidth, resulting in seven
signals, and via matricing L*,a*,b* coordinates can be estimated. An alternative approach
is to use a monochromatic sensor and a series of different coloured light-emitting diodes
(LED’s) arranged in a mosaic forming 45° circumferential illumination. Each type of LED
can be turned on one-at-a-time to generate a sequence of signals which are detected at 0°
and recorded. A matrix converts the multi-signal into colorimetric coordinates.
2
Radiant flux incident per unit area (ASTM 284 E)
3
Radiant flux emitted per unit projected area of a source (ASTM 284 E)
32
Imaging system
Another interesting group of capturing devices to record colours are colour scanners and
digital colour cameras. Although most of these three-signal imaging systems do not have
spectral sensitivities to obtain directly colorimetric values, models are available to estimate
device independent values from device dependent RGB camera response taken from the
coloured sample. For instance Brydges et al. [22] and Hong et al. [69] used a CCD colour
camera for estimating colorimetric and densitometric measurements based on polynomial
modelling. Even though the model performs well in terms of the accuracy, it is dependent
upon both the media and colorant. In other words, applying the model to other
media/colorant combinations can lead to serious eye-camera metamerism problems [140].
Instead of predicting CIELAB values from camera RGB the spectral distribution can be
estimated. This approach offers different types of post-processing, e.g. simulating the
colour coordinates using different illuminations. A study by Solli et al. [173] predicts the
reflectance spectra from camera RGB measurements. More sophisticated models to
estimate spectral reflectance from trichromatic camera systems are proposed by Imai et al.
[74]. In a recent study by Seymour [164] the author provides a review of technical papers
on the subject of the accuracy of the camera RGB to CIELAB transformation. According
to his findings, the accuracy reported is still considerably worse than one would like from a
colour measurement device. However, he is concerned about the accuracy of reporting
changes in colour within a printing process. He is proposing the use of CIELAB-like
measurements with the camera taken from a print that is considered as the reference.
Eventually the deviations from these initial values could then be used for process control.
Photometer
Light sources radiate energy in the form of electromagnetic waves. Photometry is the
science of the measurement of light, with respect to its perceived brightness to the human
eye. The photometric quantities are related to the corresponding radiometric quantities by
the luminous efficiency function V(λ) of the CIE Standard observer that models human
brightness sensitivity. Four basic photometric quantities are defined, namely the luminous
flux, luminous intensity, illuminance, and luminance. The appropriate definitions of the
terms and units are provided by ASTM E 284 [4] and ISO 3664 [79].
Luminous flux: ‘Flux’ as the light energy and luminous flux as the measure of the
flow of light energy emitted by a source, or received by a surface. The quantity is
33
derived from the radiant flux, W (in Watts), by evaluating the radiation in
accordance with the relative luminous efficiency function V(λ) of the CIE Standard
observer. The unit is lumen (lm). lm = 683 x W x V(λ)
Luminous intensity: Expresses the power of a light source. It is defined as the

quantity of luminous flux emitted in a given direction per solid angle (in steradian).
The unit is candela (cd). 1 cd = 1 lumen per steradian.
Illuminance: Illuminance (E) is a measure of the concentration of luminous flux

falling upon a surface and is expressed in lumens per unit area. The unit is lux (lx).
1 lx = 1 lumen per square meter (lm/m2).
Luminance: Luminance (L) is a measure of the flux emitted from, or reflected by,
a relatively flat and uniform surface. The unit is candelas per square meter (cd/m2)
and is also known as photometric brightness.
For in-depth information on radiometric and photometric terms and units we recommend
the reader to review Hunt’s book Measuring Colour, Appendix 1 [70].
Essentially, a photometer instrument has the same geometric specification as a
spectroradiometer and measures either the illuminance E or luminance L of light sources. It
is built similar to a colorimeter except that it only has a single channel. The tristimulus
value Y of a colorimeter reading is matching the y ( λ) function of the 1931 standard
observer and by definition reads E and L.
34
In the field of printing process control further measurement instruments have to be briefly
mentioned and addressed, measuring densities, dot area and gloss.
Optical densitometer
The optical density is the degree to which reflective surfaces absorb light, or transparent
surfaces allow the light to pass. The density is given logarithmically by,
Ii
Dink = log10 (12)
Im
where Im is the reflected light intensity and Ii is the intensity of the incident light [15].
High-density values correspond to high light absorptions. Density (absorbance)
measurements can be performed by the use of densitometers and it is often applied to
control colours in the printing process. A reflection densitometer consists of a light source
to illuminate the sample, optics to focus the light, filters to define the spectral response of
the sample and a detector to monitor the reflected light from the surface of the sample. The
sample is viewed at 45° from the surface and the reflected light is converted to density
with a logarithmic amplifier and displayed digitally.
Murray [127] expressed the relationship between the reflection density of halftone
prints and the dot area R, known as the Murray-Davies equation:
1 −10−D R
R= x100% (13)
1 −10−D H
where DR is the density of the sample and DH is the solid ink density.
Even though the revision of the ISO standard 12647-2 for process control for the
production of half-tone colour separations for offset printing processes [85] is focusing
mostly on colorimetric parameters, the density control still has a minor part of the standard
as quoted from ISO 12647-2 (section 4.3.2.3, Note 4): “Density values can be very
valuable for process control during a print run, where the instrument, the ink and the print
substrate remain the same; see ISO 13656 [88]. However, in a general situation, density
values do not define a colour to the required degree. Therefore, for the purpose of this part
of ISO 12647, reflection density values are only recommended for the determination of
tone values.” For newspaper printing the ISO 12647-3, Table A.1 of the informative
Annex A [86] gives reflection densities values provided for the solid process colours on
newsprint. It is still very common in the printing industry that the press operator first is
calibrating the solid colours according to the target values, consequently the densities
obtained with the instrument from the OK print are recorded and the densities are then
35
used as target values for process control during the production run. Although colorimetry
makes its way into the pressroom and the revised standards specify the range of acceptable
colour of the four process solids in terms of ΔE*ab, rather than in the traditional density
tolerances optical densitometers are still very often used in daily production. A study by
Seymour [165] considered the relation between ΔE*ab and Delta density, and concluded
that densitometry and colorimetry are equivalent in terms of maintaining consistent colour
on press.
Electronic planimeter
In planimetric measurements, the dot area coverage is measured by using instruments,
including the basic components of a microscope and a CCD imaging sensor. Such devices,
designed for measuring printing plates, are often called dot meters. According to Romano
[158] the major variables in such instruments are the image capturing system, aperture
selection and thresholding process. The dot meter analyzes the digital image according to
the mage histogram and decides what area coverage is a part of the dot based on a certain
defined threshold. The dot meter is actually measuring the dot area by taking a camera
snap-shot and provides an absolute value of dot coverage depending on the counts of the
number of black and white pixels in the image [34]. Research work has been done by
Wroldsen et al. [193, 194] (Paper E) to investigate whether it was possible to estimate
halftone values measured by a densitometer, from the halftone values measured by
different dot meters. Due to the poor repeatability performance for dot meters applied to
newspaper print the applied model did perform rather poorly. Hence the authors confirmed
the use of dot meters for measuring printing plates only.
Glossmeter
Gloss is an optical property of a surface and is characterized by its ability to reflect light.
The reflected light intensity from a sample can be measured by a glossmeter. The
measured intensity is dependent on the material and the angle of the illumination. If a light
beam strikes a non-metallic surface (coatings or plastics), the amount of reflected light
increases with the increase of the illumination angle. Part of the illuminated light will
penetrate into the layer of paint and the remaining light will be reflected. The resulting
amount of reflected light from a sample is given in relation to that reflected from a black
glass calibration standard with a defined refractive index. The gloss value of the reference
standard is defined to be 100 gloss units (GU).
36
Depending on the application different illumination angles are used. The gloss
characteristics of different paper type substrates are classified in ISO 12647-2 and are in
accordance with ISO 8254-1, TAPPI method 75° [80].
2.2.2 Sources of Error and Measurement Uncertainty

When selecting a colour measurement instrument a number of factors might be considered.
Most of the applications require spectral data, which provide information about the
properties of raw materials, content of fluorescent whitening agents (FWA), potential
problems with metamerism, and the ability to compute colorimetric data independently of
light source. Furthermore the selection of the measurement conditions (radiance,
irradiance) and the appropriate geometry is significantly determining the measurements
[16, 151]. The suitable software to calculate and read out the measurements, service, and
support agreements are further important factors to consider.
An additional aspect is the question about the reliability of the measurement data
obtained from a certain sample and the procedure to minimize the measurement
uncertainty. Uncertainty of measurement is an estimation of the dispersion of the values
within which the true value lies [138]. It is known that two models of spectrophotometer
will produce different results from measurements on the same sample due to different
instrument design in the area of the instrument’s measurement system including the
measurement geometry, the aperture size, measure position and filter used, light source,
dispersion function, the scanning interval and the bandwidth. Depending on the properties
of the measured samples (texture, gloss or FWA) these differences in the measurement
system of the instrument can affect the measurements [14].
The instrument calibration and verification procedure is contributing to reducing
the source of error as demonstrated by Briggs et al. [21] where the authors discuss
reliability issues for colour measurement in quality control applications. They conclude
that the measurement accuracy is an important aspect of any system for colour
measurement and that the calibration procedures can significantly reduce colour
differences. A study by Zwinkels [200] demonstrates the steps in calibrating and verifying
the performance of colour measurement instruments. Furthermore, the author emphasizes
the importance of specifying the procedures of spectrophotometric measurement on
surface colours with 1) the quantity measured (reflectance or transmittance), 2) the
geometry of illumination and viewing, 3) backing material used, 4) aperture size, and 5)
the white reflectance standard for instrument calibration and its traceability. Berns [16]
37
defines the calibration as the process of adjusting an instrument in such a way that its
readings reproduce a national or international scale. Furthermore, the verification is the
process of assessing an instrument’s ability to precisely and accurately reproduce, national
or international measurement scales.
According to Berns [16] measurement uncertainty can be divided into precision and
accuracy. Precision describes the dispersion of the measurements taken and accuracy
describes the conformance of a series of measurements to the accepted or true value.
Precision can be further divided into repeatability and reproducibility. ASTM E2214 [9]
specifies repeatability as a measure of how well an instrument repeats its reading of the
same sample over a certain period of time (e.g. short-term, medium-term, and long-term
repeatability). The instrument’s repeatability is quantified by calculating the colour
difference between the mean of the measurements and each individual measurement taken
(mean colour difference from the mean (MCDM)).
MCDM = ∑ ΔE i /N (14)
where ΔEi is the colour difference between each measurement and the mean of the
measurements and N is the number of measurements. Reproducibility is a form of
repeatability in which one or more of the measurement parameters have been
systematically changed such as the sample is different, the time frame of measurements are
very long, the procedures or instrument are different or the operator has changed. A report
by Scorer et al. [163] defines procedures to be used to monitor the performance of an
instrument in terms of its repeatability and reproducibility.
Inter-instrument agreement describes the reproducibility of two or more
instruments of identical design. On the other hand inter-model agreement expresses the
reproducibility of two or more instruments of different design. In other words,
reproducibility determines the variations between instrument’s readings. In instruments,
which have the same design, the random amount of bias is reduced compared to
instruments with different design. A comprehensive set of works was carried out by Wyble
and Rich addressing the evaluation of methods for verifying the performance of colour
measurement instrument. Part I is addressing the repeatability issue [196] and part II is
covering inter-instrument reproducibility [197].
Accuracy is affected by systematic errors, which are characterized by the fact that
they remain fixed when the measurement is repeated under the same conditions. In a report
published by the National Physical Laboratory (NPL) [33] spectrophotometric errors
38
affecting the measurement results are discussed. According to the author, possible
uncertainty sources are defined such as:
• Scale uncertainty (usually via instrument reference standard, e.g. white tile
standard)
• Photometric linearity
• Dark level uncertainty
• Wavelength uncertainty
• Bandwidth uncertainty
• Gloss trap or specular beam uncertainty, for integrating sphere geometries
• Repeatability uncertainty
It is beyond the scope of this introduction to discuss in-depth the details of these
systematic errors. Therefore we refer to the valuable references [33, 186, 200]. A work by
Robertsen [155] is addressing the issues of systematic errors and inter-instrumental
agreement and proposes a method to assess the severity of these errors. Berns and Petersen
[17] have further redefined and extended the proposed method. In another study by Rich et
al. [152] maximum colour differences of up to 4.0 CIELAB units were observed between
pairs of instruments. Similar pairwise colour differences were reported by Nussbaum et al.
[136] (Paper F) on BCRA tiles. But on paper substrate the inter-instrument agreement was
even higher in terms of colour differences (ΔE*ab > 6). Another source of error can be
thermochromism which is the change of colour with changing temperature. In fact, the
instrument’s light source should not cause the sample to change temperature [33]. A study
by Fairchild and Grum [43] demonstrates the effect on highly coloured samples, such as
red, orange or yellow having spectral distributions with steep slopes, and propose a
correction method. Various other studies and research addressing measurement
uncertainties and comparative studies of colour measurement instrumentations have been
presented (see e.g. Billmeyer [18, 19] and Rodgers et al. [157]).
Several attempts have been made to correct the measurement data from different
instrument readings applying mathematical models to improve the inter-instrumental
agreement. De Garcia et al. [35] evaluate the colorimetric behaviour of different
spectrophotometers. A study by Steder et al. [174] is assessing the selection of training
data in terms of which and how many samples are needed for training the model while
maintaining a good performance. Correction models applied to the spectral data have been
proposed by Berns and Petersen [17] and Chung et al. [24] to improve the inter-
instrumental agreement on textile. Nussbaum et al. [133] (Paper G) demonstrate a
39
correction model applied to CIELAB measurement data set to improve the colorimetric
performance and hence the inter-instrument and inter-model agreement in printing
applications.
For measurement and calculation procedures for self-luminous video display
devices, such as CRTs and flat-panel displays, other standards or recommendations are
provided including CIE 122 [27], the IEC 61966-2-5 [73], and the ASTM standards E1336
[6] and E1455 [8]. These specifications present measurement procedures as well as
measurement instrument characteristics in particular for spectroradiometers and tristimulus
colorimeters.
40
Background Image Reproduction and Colour Management
2.3 Image Reproduction and Colour Management

Having reviewed the colour fundamentals, and discussed methods and instruments for
measuring colour and light, this section is aiming to give an overview on image
reproduction and colour management. A wide range of comprehensive literature and
textbooks providing details on image reproduction and colour management exist and the
reader is referred to literature of e.g. Giorgianni and Madden [56], Green [58, 60], Morovic
[121], G. Sharma [169], A. Sharma [167], and Adams et al. [2].
2.3.1 Additive or Subtractive Colour Mixing

Additive colour systems produce colour on a dark background by the combination (adding)
of different wavelengths, known as primary colours. The term additive is used to signify
the fact that the final spectrum is the weighted sum of the spectra of the individual lights.
Typically, the additive primaries are red, green, and blue (RGB) and the combination of
red and green forms yellow, red and blue forms magenta, and blue and green forms cyan as
illustrated in Figure 12 a). The combination of all three primaries at full intensities
produces white and the intermediate colours are obtained by varying the individual
primary intensities.
In contrast, subtractive colour reproduction is used for transparent and reflective
media, and is produced by removing (subtracting) unwanted spectral components from
“white” light. By deposing different colorants onto these media different colours are
generated by selectively absorbing light of certain wavelengths while transmitting other
wavelengths. The most common subtractive colorants are based on cyan, magenta, and
yellow (CMY) that absorb light in the corresponding spectral regions of red, green, and
blue, respectively. Roughly speaking, each colorant absorbs its complementary colour and
transmits the rest of the visible range of the spectrum. In other words, the individual CMY
colorants eliminate RGB spectral regions, respectively, as illustrated in Figure 12 b). The
combination (overlay) of cyan and magenta eliminates both red and green, producing blue;
the mixture of cyan and yellow eliminates red and blue, producing green; and the
combination of magenta and yellow eliminates green and blue, producing red. The
combination of the maximum amounts of all three produces black, and by varying the
colorant amounts the intermediate colours are produced.
41
Image Reproduction and Colour Management Background
K W
R C
Y M B
W K G
G C B M R Y
a) b)
Figure 12: a) The additive mixing of RGB primaries and b) subtractive colour mixing of
CMY.
It is common to add a fourth black (K) colorant in subtractive systems to improve the
reproduction of achromatic (grey) colours and to extend the gamut by producing darker
colours. Halftone colour printing can be considered as a hybrid system because the
colorants combine subtractively, but the perceived colour is the average of the differently
colored regions over a small area [170].
2.3.2 Image Reproduction

In the context of image reproduction the original image can originate from multiple
sources. Images of real objects in real scenes exist as spatial variations of spectral
distribution of radiance and reflectance. To record the real colour information, the scene
has to be captured with a colour reproduction device (medium) such as a digital camera, by
sampling the colour information both spatially and spectrally. Consequently, the recorded
image can be processed and reproduced by using additive or subtractive colour mixing
with a particular set of primary colours. Before commencing this introductory discussion
on the principles of image reproduction and colour management it is worth clarifying some
issues of terminology. In the CIE 156 publication [30] “Guidelines for the evaluation of
gamut mapping algorithms”, the key components of image reproduction and their terms
for image, colour reproduction medium, colour gamut, and colour gamut mapping are
defined as follow:
Image: “Two-dimensional stimulus containing pictoral or graphical information

whereby the original image is the image to which its reproduction is compared in
terms of some characteristics (e.g. accuracy).”
Colour reproduction medium: “A medium for displaying or capturing colour

information, e.g. a monitor display, a digital camera or a scanner. Note, that in the
42
case of printing, the colour reproduction medium is not the printer but the
combination printer, colorants and substrate.”
Colour gamut: “A range of colours achievable on a given colour reproduction

medium (or present in an image on that medium) under a given set of viewing
conditions – it is a volume in colour space.” The colour gamut of an image or
colour reproduction medium can be represented in a three-dimensional color space
such as CIEXYZ or CIELAB.
Colour gamut mapping: “A method for assigning colour from the reproduction
medium to colours from the original medium or image (i.e. a mapping in colour
space).”
Figure 13 illustrates the key components of the image reproduction process including two
colour reproduction media or devices with their corresponding colour gamut. The input
colour reproduction media are those capturing digital colour information from a visual
stimulus, either a real scene or an object. Scanners and digital cameras are typically
considered as input media reporting RGB values for each of the pixels it records. The
values represent the amount of light detected at a given stimulus location through the
device’s RGB filters. Since the RGB values recorded by an input device have a
colorimetric meaning that is a function of the device’s photodetector properties, the filter
RGB’s of each individual camera end up having somewhat different colour meanings.
On the other hand, output media take a two-dimensional image array of digital
values and generate a stimulus as a result. In case of a printed medium as an output the
printer deposits device dependent CMYK colorants on a substrate based on digital inputs
and produce a subtractive colour reproduction. Technologies for depositing colorants on
substrates include dye sublimation, thermal wax, inkjet, laser, colour photographic prints,
and halftone colour printers. All these are representatives of using the subtractive colour
mixing process.
The key property of device dependent colour spaces is that they do not in
themselves have colorimetric meaning. For example, a device CMYK of [0%, 10%, 20%,
…100%] does not give any information about colorimetry or colour appearance, it is just
an instruction to the printer engine how much colorant it has to deposit onto the substrate.
Device colour spaces also map onto the colour range of the device that they are used with,
which makes them particularly suitable for controlling device outputs. The appearance of a
final print is a function of the printing device’s characteristics (e.g. tone reproduction
43
curve), the primary colour used, the properties of the substrate and the viewing conditions
under which the print is viewed.
There are other types of output colour reproduction media to consider. These are
displays using different technologies such as e.g. cathod-ray tubes (CRT), liquid-crystal
display (LCD), organic light emitting diode displays (OLED), plasma displays and digital
micro-mirror devices (DMD) respectively. The display media use mainly additive colour
reproduction. However, the colour appearance is strongly dependent on the viewing
conditions [121].
Device specific colour models means that equal sets of RGB or CMYK numbers
will produce different colours (different colorimetric meaning) on different devices (or
different substrates on a printer or a printing press). To produce the same colour on
different devices the RGB and CMYK numbers need to be changed and adjusted to each
device. Hence, RGB and CMYK are just instructions for a device of how much colorant to
use (e.g. to produce equal colours on different devices).
L* L*
-a* -a*
-b* +b* -b* +b*
+a* +a*
Original Image
Colour Gamut Mapping
Image Reproduction
Input Output
Colour Reproduction Colour Reproduction
Medium Medium
Figure 13: Schematic overview of the key components of image reproduction.
44
2.3.3 Colour Gamut Mapping

A key part in image reproduction is colour gamut mapping, which is to transform colour
information between colour spaces. In other words, gamut mapping provides the
connection between the original’s appearance (source) and the appearances possible in the
reproduction (destination) and has to map each colour in the source gamut to a colour in
the destination gamut. In order to be able to reproduce the original image properly,
modifications in some colours often have to be done which either affects the accuracy or
the pleasantness of the reproduction. The substitution of original colours in the image
source colour space by colours which are reproducible in the destination colour space can
often lead to loss of information caused by shape deformation and size reduction of the
colour gamut between an original image and its reproduction. To overcome these
challenges an impressive number of different gamut mapping algorithms (GMAs) have
been proposed in the literature. Morovic and Luo have made a comprehensive survey in
[121, 123, 124]. The authors classified the point-wise GMAs into two categories, gamut
compression and gamut clipping, respectively. For further details considering the two
reproduction intents we refer to Morovic and Luo [125].
For a long time, most of the research activities on GMAs have been related to
simple mappings in colour space. A different concept of performing gamut mapping is to
take the spatial dimension into account. Farup et al. [46] present a review of different types
of spatial GMAs and their characteristics. Further work in perceptual evaluation assessing
the performance and gamut mapping quality of different type of GMAs has been done by
Bakke et al. [10], Bando et al. [12], Bonnier et al. [20], Dugay et al. [37], and Morovic
[121].
The construction of a gamut boundary descriptor (GBD) is the first step in
performing a colour gamut mapping transformation. The performance of a GMA depends
on the used GBD estimating the appropriate gamut volume. Bakke et al. [11] introduced a
method for GBD evaluation; among a number of GBD’s tested the authors conclude that
the modified convex hull algorithm performs well on a range of different data sets.
45
After reviewing the fundamentals of colour image reproduction let’s focus on the
principles of ICC colour management for print production.
2.3.4 Principles of ICC Colour Management

The basic aim of colour management (in the graphic arts industry) is to ensure colour
accuracy throughout the entire workflow from initial draft through the finished printed
product. To achieve this goal it is necessary to map the colour reproduction characteristics
of each input and output colour reproduction medium within the device independent colour
space (CIEXYZ or CIELAB) and to store this information in ICC colour profiles. Thus,
the final colour of the print can be simulated at any stage in the production workflow [68].
Colour management as such is no different from the process of colour reproduction used in
the graphic arts industry for many years except moving from a closed-loop colour
management system to an open architecture using a device independent colour
reproduction system where each source and destination is being handled independently.
To be able to adopt and implement this new architecture managing colour in the
printing industry the International Color Consortium (ICC)4 was established in 1993 by
eight industry vendors “for the purpose of creating, promoting and encouraging the
standardization and evolution of an open, vendor-neutral, cross-platform colour
management architecture and components” [178]. The development of ICC profile
specification with the current version called version 4 (v4) profiles [72] was the result of
this co-operation. The ICC profile specification is also standardized and published in ISO
15076-1 [90] which specifies a colour profile format and describes the architecture within
which it can operate. The key components in the ICC architecture are profile connection
space (PCS), ICC profile, rendering intents (RI), and colour matching module (CMM).
Essentially the solution proposed by the ICC is based on a process where the colour
reproduction is divided into two transformations. The forward colour transformation takes
the device dependent colour information (e.g. from a camera or a scanner) and transforms
it into a device independent colour space (colorimetric), called profile connection space
(PCS). Then, the inverse colour transformation takes the colorimetric information and
transforms it into a device colour space such as RGB or CMYK (see Figure 14). The ICC
device profile for a given colour reproduction medium describes the relationship between
device control signals (RGB or CMYK) and the actual colour those signals predict or
produce. It defines the device independent values (CIEXYZ or CIELAB) that correspond
4
International Color Consortium (ICC), http://www.color.org
46
to a given set of device dependent numbers and vice versa. ICC [72] provides detailed
specifications of the structure of a profile including their internal operations and the
specification of the architecture.
Note that each colour reproduction medium is required to have its particular device
profile rather than just one for each imaging device. For example a printing machine will
need different profiles for different type of substrate (e.g. glossy paper and recycling
paper).
All colour information is transformed through the PCS combining ICC source
profile and ICC destination profile. (Here the term “source-destination” is used instead of
“original-reproduction”. However, the meaning of these pairs is the same.) The PCS is
based on the colorimetric colour space CIEXYZ or CIELAB determined for a specific
observer (CIE Standard 1931 Colorimetric Observer), relative to a specific illuminant
(D50), and measured with specified measurement geometry for reflecting media (0°/45° or
45°/0°). Measurement procedures are also defined for transmitting media. In the current
specification, two variations of PCS are defined. One is an original-referred variation for
colorimetric intent profiles, and the other is a standard output-referred variation for
perceptual intent profiles where the measured data was not made relative to D50. The
profile builder application is expected to correct the data to achieve this (derived from ICC
[72]).
As discussed above, each colour reproduction medium (device) has a fixed range of
colours that it can reproduce under certain viewing conditions, called, its colour gamut.
Colours that are represented in the source colour space but are not reproducible in the
destination colour space are called out-of-gamut colours. Consequently those colours have
to be replaced by other colours in the destination colour space. The ICC framework
proposes four different types of colour transformation methods called rendering intents
(RI) to allow the user to select the one appropriate for the desired purpose. The rendering
intents constitute a permanent component of the ICC profile.
Media-Relative Colorimetric Intent: Generally, the colorimetric rendering intents

permit within gamut colours to be reproduced accurately. It rescales the in-gamut
colours in such a way that the medium's white point is mapped to the PCS white
point (for either input or output). It is useful for reproduction in particular for
professional photography where the print is similar to the original viewed on a
display.
47
ICC-Absolute Colorimetric Intent: The tristimulus values of the in-gamut

colours remain unchanged by using the ICC-Absolute Colorimetric Intent. It is
typically useful for artwork reproduction where the transformation is intended to
reproduce the original appearance in terms of the illumination and the substrate.
Furthermore, a very important application in the graphic arts industry is the
simulation of one medium on another (proofing).
Perceptual Intent: This is intended for preferred or pleasing reproduction of

images, particularly pictorial or photographic-type images. It is especially useful
when the source and destination media are substantially different and a pleasing
colour output is desired. For example if the destination medium has a smaller
colour gamut than the source medium, perceptual rendering may alter in-gamut
colours to allow gamut compression. If the destination medium allows for greater
chroma than the source medium, then chroma may be increased to produce a more
pleasing result. Since perceptual renderings are vendor specific the results may
differ between profiles produced by different vendors. Therefore users have to be
aware of this by transmitting output profiles along with their images if a consistent
workflow is required as in distributed printing.
Saturation Intent: The saturation intent is also a vendor specific gamut mapping
transformation. In order to preserve the vividness of pure colours it involves
compromises, such as trading off preservation of hue. It is useful for rendering
images, such as business graphics (pie and bar charts), where we simply want vivid
colours and aren't particularly concerned with reproducing the colours exactly. In
the graphic arts industry the saturation rendering intent is the least used in day-to-
day practice.
The choice of the appropriate rendering intent can have a significant effect on the colour
reproduction. Therefore the selection is an important decision for the user to determine the
aim of the reproduction.
A colour management module (CMM) performs the actual conversion of the colour
data from one colour space to another with the aid of colour profiles. A CMM performs the
actual conversion of the colour data from one colour space to another with the aid of ICC
profiles. An ICC profile contains a) device colour to PCS translation parameters and b)
PCS to device colour translation parameters.
48
Typically the communication between the PCS and input device profiles is one-way,
because the PCS needs to know what colours the input RGB numbers represent (e.g. from
a digital camera). On the other hand, monitor display ICC profiles must be two-way,
because the display acts as both an input and an output device (Figure 14). If you create or
edit a colour image based on its appearance on your monitor display, you are using your
monitor as an input device. If you display an image on your screen, the monitor is the
output device. Printer profiles are always two-way profiles, too. We use them not only to
convert them from the PCS to the output colour space for printing, but also to display files
already converted to output space on the monitor display, as discussed in the study of Sole
et al. [172] (Paper C) considering soft proofing, or to convert a press CMYK image to
some other output device’s space for proofing [50, 167].
Display Digital Printer

ICC profile ICC profile
and RI and RI
Scanner
ICC profile
and RI
RGB
RG YK
B CM
Offset press
Camera ICC profile
ICC profile PCS CMYK and RI
and RI
RGB
Figure 14: Overview of ICC workflow including device profiles via the PCS and rendering
intent (RI) choices.
For the procedure and creation of ICC device profiles and issues addressing assigning and
converting source profiles to destination profiles including suggestions of different types of
colour workflows (e.g. RGB or CMYK) we refer to textbooks focusing on practical colour
management application from Homann [68], Fraser et al. [50], Sharma [167], and
Johansson et al. [98].
2.3.5 Calibration and Characterization

As we have seen above, device profiles describe the relationship between device dependent
colour signals (RGB or CMYK) and the corresponding device independent colour output
signals (CIEXYZ or CIELAB) under certain given calibration.
49
Johnson [99] defines calibration as an operation of establishing fixed, repeatable

conditions for a certain device or process to always produce the same colour according to a
given set of numbers. The measured values agree with the values specified by a standard or
a characterization process and it is an essential first step before any colour imaging process
takes place. On the other hand, characterization simply defines the process of relating
device-dependent colour (e.g. CMYK values) to device-independent colour values (colours
measured on the printed sheet) according to a given printing definition. For example using
a colour printing system the characterization process defines the relationship between a
digital CMYK colour data set and the resultant CIE colour measurement data obtained.
ISO 12642-2 [82] defined an expanded input data set for characterization of 4-colour
process printing represented either with the 928 patch IT8.7/3 target or the 1617 patch
IT8.7/4 target and the corresponding description of its associated printing definition.
Recently, another standard defining “Methods of adjustment of the colour reproduction of
a printing system to match a set of characterization data” has been approved in ISO 10128
[97]. In order to retain the validity of a characterization, it is necessary to re-calibrate the
image reproduction devices (such as a monitor display, digital printer, or conventional
offset press) at regular intervals. From any type of characterization different profiles with
different properties can be created.
2.3.6 Standardization in Offset Printing and Prepress

The characterization of traditional printing processes is much more complex and time
consuming compared to the generation of ICC profiles for scanners, monitors, and digital
colour printer systems. The colour appearance of a printed product is affected by a number
of analogue and digital sub-processes, and their respective process parameters applied. The
production processes themselves including the raster image processor and digital plate
production, the materials used, such as paper substrate type and inks, the printing process
used (including all its specific technical characteristics), and operator dependent variables,
such as the quantity of ink applied, the ink/water balance, all affect the appearance of the
colour in some measure. Obviously the standardization of all these variables is
indispensable for the successful use of a colour management system in a professional
production workflow. All these factors and parameters are implicitly included in an ICC
colour printer profile established for a particular printing process. At the very least, in-
house procedures must ensure a stable production sequence, otherwise the validity of the
ICC profiles generated for the printing process will be compromised.
50
Essentially, there are two different approaches considering printing press

conformance, namely optimized or standardized press behaviour. A fully optimized press
aims at maximizing its capability in terms of lowest possible dot gain, highest ink densities
and best contrast that the individual printing press can achieve, without considerations of
any external specifications or standards. Such individual parameters can create unique
press conditions that require custom ICC profiles for creating the appropriate separations.
Initially, after the new ICC colour management architecture was introduced in the printing
industry and commercialized profiling tools were available it was very common to
generate custom printer profiles. Very often, such custom printer profiles were related to a
very small number of test prints and colour measurements representing the print
conditions. Together with inconsistency of the process calibration in terms of absent
repeatability and uniformity specifications, the performance of these profiles and the
corresponding print quality has been considered as not satisfactory. Moreover the
variations between printing presses and the corresponding conditions were rather large,
caused by inconsistency of using common target values of the solid primary colours and
common dot gain convention even though the same type of printing substrates had been
used [130-132] (Paper B and Paper D).
Another approach is to make the presses conform to a certain reference or standard
such as e.g. ISO 12647-2, and applying predefined parameters. ISO 12647-2 specifies a
number of process parameters and tolerances including the corresponding values to be
applied when preparing colour separations for four-colour offset printing [85]. Similar
process parameters are specified in ISO 12647-3 for coldset offset lithography on
newsprint [86]. By using the second approach and by standardising the behaviour of the
press, industry standard ICC profiles can be used. The European Color Initiative (ECI)5 is
providing ICC output profiles for offset and other printing technologies based on reference
characterization data according to the ISO 12647 series of standards [84-87]. Any single
printer output profile contains a specific set of parameters for different separation
elements; a single level of grey component replacement (GCR), one certain colour
separation, one total-dot-area setting, one method of gamut compression, one tone
reproduction curve, and so on. Hence, different output profiles contain different
combinations of these separation parameters and thus provide multiple options to adapt
data to a particular set of characterization data [58].
5
http://www.eci.org/doku.php?id=en:start
51
In terms of aiming for a common print appearance across printing plants (e.g.
preserving print appearance of an ad campaign) the second approach including consistency
will be the most suitable one to ensure a predictable and equivalent print result. However, a
constant maintenance of the calibration including using the appropriate parameters for the
solid primary colours, dot gain and grey balance is a very important requirement for
predictable print quality. Methods and procedures for print quality assessment and process
control according to objective evaluation and perceptive judgement including appropriate
viewing conditions according to ISO 3664 [79] are proposed by Nussbaum and Hardeberg
[131] (Paper D) and presented in Part II in this thesis. The demand for offset printing
standardization in terms of using international print parameters and the exchange of
colour-managed data is increasing – with corresponding effects on how modern print
enterprises work. International standards and Process-Standard Offset (PSO)6 offer
predictable results and consistent, comparable quality location and output device
independent.
To reduce expensive and time consuming iterations in standardized printing
workflow soft proofing has become an important method in predicting the final print
product on a monitor display. ISO 12647 [83] defines soft proofing as “the ability to match
colour images displayed on colour monitors to the images produced when the same digital
file is rendered by proofing and printing systems”. In the past a number of studies and
research work have addressed the issue of soft proofing. For more details we refer to the
work of Gatt et al. [53, 54], Hardeberg et al. [65], and Roch et al. [156]. However,
although the concept of soft proofing is not new, in practical applications the colour
appearance between two different media (e.g. softcopy simulation of a hardcopy) can
differ a lot due to unsuitable type of devices, viewing conditions, incorrect use of standards
and parameters, inaccurate device calibration and characterization, or inappropriate
measurement methods. Although ISO 12646 defines parameters for monitor and viewing
booth condition setup for soft proofing environment, the practical methods to implement
these standards and parameters as per the job requirements have not been clearly defined.
Consequently, a mismatch between the viewed stimulus on the monitor display and the
corresponding stimulus on printed substrate in the viewing booth can occur. The paper by
Sole et al. [172] (Paper C) is aiming to describe in detail how to set up an appropriate soft
6
PSO assures predictable quality print production from data creation to the final print production.
Industrially orientated and standardized method for the creation of print products. http://www.pso-
insider.de/index.php?id=7&L=1 and http://colorlab.no/psoinfo
52
proofing work station (comprising of a monitor and a viewing booth with the appropriate
ambient lighting conditions) according to the parameters given by ISO 12646 for soft copy
and hard copy proof comparison in the graphic arts industry, in order to evaluate the
performance of the entire soft proof setup according to the values and tolerances defined in
ISO 12646. This is also in accordance with soft proofing in a standardized printing
workflow according to PSO.
In print standardization and Process-Standard Offset (PSO) certification
requirements, viewing conditions in the prepress and pressroom have to be measured and
verified according to the parameters defined in the ISO standards [79, 83]. Consequently,
appropriate measurement instruments are required, and often spectrophotometers and
colorimeters are used to determine the photometric quantities (illuminance and luminance
values) in the production workflow. Most of the commercial measurement instrument
manufactures supply the instruments with an ambient light measurement adaptor to be
placed in front of the measure aperture.
2.3.6.1 Process-Standard Offset (PSO)

PSO is an industrially orientated method or standardized procedure to control the various
steps in the creation of print products. The idea of PSO was developed by the The German
Printing and Media Industries Federation7, in collaboration with Fogra8 and UGRA9 and is
in conformance with the international standardization series ISO 12647. The
implementation of PSO assures predictable quality print production from data creation to
the final print production following the functions:
• Organization and documentation ISO 9000 [81]
• Data reception and data creation ISO 15930-x [91-94]
• Display/Soft proofing ISO 12646
• Proofing ISO 12647-7
• Standard illumination ISO 3664
• Plate making ISO 9000
• Print production ISO 12647-2
In PSO, testing devices and control methods are described with which the production
process can be controlled. It describes the input to output workflow, which includes a
number of ISO standards describing this workflow.
7
http://www.bvdm-online.de/Aktuelles/
8
Fogra Graphic Technology Research Association, http://www.fogra.org/
9
Association for the Promotion of Research in the Graphic Arts Industry, http://www.ugra.ch/
53
54
Background Print Quality and Print Assessment
2.4 Print Quality and Print Assessment

The previous section discussed the fundamentals of image reproduction and colour
management including printing standards and methods. This section is focusing on print
quality, discussing various factors affecting the print quality and the use of different
assessment methods and procedures to determine the print quality. Beside the introduction
given in the thesis a considerable amount of literature is devoted to image quality and
assessment methods. In particular, the reader is referred to the textbooks by Gescheider
[55] and Fairchild [42] for the fundamentals of psychophysics, Keelan [103] for
characterization and production of image quality and Engeldrum [38] who proposes an
image quality toolkit for imaging systems. Hunt [71] explains the visual appreciation and
the basis of judgement regarding colour reproduction. Kipphan [105] discusses factors for
print quality assessment and specifications for process control. Different aspects of colour
printing quality are proposed by Field [48]. The textbook by Wang and Bovik [189]
provide details about objective image quality assessment including a broad treatment of the
current state-of-the-art and future directions.
Since the human observer is the ultimate receiver in most image reproduction
applications, the most reliable way of assessing the quality of an image, a print or a
document is by subjective evaluation. On the other hand, depending on the type of
application, subjective assessment methods are not preferred or appropriate due to
significant individual differences in the observer judgements. Thus, objective methods
should be applied, attempting to simulate the functional components in the human visual
system.
How can we classify image quality, print quality, and document quality? Before we
discuss the subject of this section let’s briefly define the three types of measures in the
context of the present work. Even though there is no distinct borderline between image,
print, and document quality assessment we can address them as different measures. Often,
image quality is related to digital images which are subject to a range of distortions, caused
by acquisition, processing, compression, and reproduction, resulting in a degradation of
visual quality [190]. According to Silverstein and Farrell [171] there is a natural tendency
to confuse image quality with image fidelity and the two terms are often used
interchangeably. The authors point out a clear distinction between the two measures and
define image fidelity as the ability to discriminate between two images and image quality
as the measure of preference for one image over another. For example an original image
55
Print Quality and Print Assessment Background
reproduced and printed in different location: if we cannot detect the differences between
the prints, we might conclude a high print fidelity. On the other hand, if an original image
is “enhanced” by a distortion, people may detect the difference between the original and
the distorted version and prefer the distorted version over the original. We may conclude
that it is either the image and its content that is determined by the quality aspect or it is the
behaviour of the reproduction medium, which is in focus. For example by assessing
different types of GMAs the image quality obtained by the different transformations is in
focus, independently of the behaviour of the reproduction medium used to view or print the
image reproduction. In contrast, assessing the performance of the reproduction medium
itself, the image quality is not necessarily of interest but the behaviour and the quality of
the device to capture, view or print a stimulus is of importance. In other words, even
though the performance of a device conforms to specification, the image quality can be
distorted due to some error in the input data. A further term, “document quality”, has been
used in other studies by Falkenstern et al. [44, 45] to indicate that images are only one
graphical element among others. Text, vector illustrations, charts, and vignettes are also
visual element integrated in a document.
If the behaviour and performance of an input or output reproduction medium has to
be evaluated then the measure of the device quality has to be considered. As the scope of
this thesis is related to printed media, the quality and control of the printing process is
relevant.
2.4.1 What is (Print) Quality?

The term print quality can have many aspects depending on the context or the way the
issue is considered. The quotes10 made by print customers regarding their printed copies
and the judged print quality illustrate the complexity in terms of quality and expectations:
“I just received my first book and thought everything was really good except the
print quality. The print shop advertised they can print bookstore quality photo
books, but I can see the printer roller marks on the page and all my images have
lost their detail and look blotchy…”
“…same problems, no grey, some of the pictures have a colour cast others are too
dark.”
“I'm happy with the print quality - the colours turned out vivid and sharp - the
pictures look close to what I see on my monitor…”
10
Quotes collected from Blurb Forums: http://forums.blurb.com/forums/5/topics/2498
56
“I calibrate my monitor once a week and, well, I know what I'm doing. I just
received my colour prints and I notice no detail in blacks, lots of print grain in the
mid-tones, blotchy skin tones, banding on every page...”
The term "quality" is often left unclear because definitions are vague or incomplete. Even
though the common definition, "quality is meeting or exceeding customers' expectations,"
is quite true, it is completely useless if those expectations are undefined. Even if print
specifications are identified in advance, and the definition "quality is conformance to
specifications" is proposed, this statement can be extremely narrow and does not
necessarily reflect what customers really mean by the term "quality." Field [48] adapted a
generic framework of product quality, proposed by Garvin [52] and applied it for the
specific needs of the traditional colour printing industry. He is considering nine aspects of
colour printing quality, divided into four appearance factors and five valued by customers.
Appearance factors:
Conformance to specifications: This factor refers to how closely the final product
conforms to pre-defined tolerances in terms of density or colour targets and
registration goals.
Technical excellence: Concerns the physical and psychophysical properties of the

colour reproduction. Tone and colour reproduction, image definition, interference
patterns and surface characteristics are the main components of this aspect of
quality. The printing conditions, the characteristic of the original, and the specific
instructions or demands influence the optimal values for these attributes.
Aesthetics: It refers to the creative decisions made by the photographer or graphic

designer to express their message in the advertising or fashion illustration business.
The printers as such do not have significant influence on the aesthetic aspect of
quality.
Permanence: It is related to the ability of the substrate and inks to resist

environmental influence of light, chemicals, and moisture. Overprint varnishes or
coatings can also influence the permanence.
The non-appearance aspects of colour printing quality are related to customer service,
production, logistic, and economics. Recently, a new quality label of the graphic arts
57
industry has been introduced. The swissPSO11 label stands for a comprehensive quality
covering the entire workflow, from initial customer contact through the production until to
the disposal of the waste.
2.4.2 Factors Contributing to Print Quality

At the highest level, printing can be seen as having digital data as its input and a visual
stimulus as its output. The relationship between the digital data and the corresponding
visual stimulus, i.e., a print viewed in a certain way, can be highly dynamic. As various
factors are involved in the printing process as well as in the viewing of the resulting print,
there is great potential for variations determining the final print quality.
Considering a generic printing workflow the starting point is a digital image, which
is sent to a printer from some software application using a certain printer driver. For non-
impact printing technology (NIP) the instructions from the printer driver are sent directly
to the imaging engine of the printing system, e.g., LED, inkjet, dye-sublimation, which
deposits colorants on a substrate. In terms of conventional printing the instructions from
the printer driver are sent via a raster imaging processor (RIP) to the computer to plate
(CTP) system where the printing plate or image carrier is the material by which ink is
transferred to the printing substrate [105]. The final print, when viewed in a certain
environment under a certain light source, results in a certain appearance for a viewer. Any
change to the properties of any of the elements involved in the workflow has the potential
to alter the final appearance corresponding to the digital input. Thus, the workflow can be
typically divided into categories and each category can have certain factors affecting the
appearance of the print as demonstrated in Figure 15. The different factors can be
categorized into two classes. Those that determine the resulting print for a given digital
input, and those that determine the colour appearance of a given print. Multiple factors and
their alternative states, (e.g., the light source can be ‘A’, ‘D50’, or ‘office lighting’)
affecting the appearance and consequently the quality of the print have been discussed in
studies by Nussbaum and Morovic [135] and Morovic and Nussbaum [126]. Although, this
reproduction process is rather complex in its nature high fidelity and consistency of the
reproduced images are expected in terms of the perceived print quality.
11
http://www.swiss-pso.ch/index.cfm?sID=720&sprache=1
58
Digital Input to Print Print to Appearance
Digital Printing Viewing

Print
Input System
• Data format • Printing technology • Uniformity • Stimulus and viewing
• Colour space • Characterization • Repeatability conditions (e.g. size,
• Resolution • Primary ink colours • Temporal stability distance, geometry, flare
• Compression • Dot gain • Colour measurement background, surround)
• Printer driver • Dot shape • Surface
• Calibration • Grey balance • Postpress Illumination
• Colour management • Register
• Light source
• RIP/CTP • Substrates
• White point
• Intensity
Figure 15: From digital input to print appearance.
Essentially, print quality can be seen as a combination of a number of individual attributes

or factors contributing to the overall perceived print quality. A work conducted by
Rasmussen [150] identifies the most significant attributes affecting the printed image
quality produced by a colour printer. He proposes different approaches to measure print
quality dependent on a specific goal (e.g. manufacturing, development), and to evaluate
which measurement techniques meet those fundamental requirements. Marcu [115, 116]
then deals with various factors that determine print quality such as printing technology,
colorant/media interaction, geometric resolution, halftoning, separation, black generation,
under colour removal (UCR), grey component replacement (GCR) and tone reproduction.
Kipphan [105] refers to factors of influence and specifications determining the quality of
print and is dividing the production process into the components “prepress”, “print”,
“postpress”, and “material”. Each component has multiple factors influencing the print
quality, similar to what we have seen in Figure 15. On the other hand the specification to
determine the print quality must be definable and measurable. Kipphan defines four
categories named “colour”, “resolution”, “register”, and “surface” and attributes attached
to them. For example colour coordinates, optical density, dot gain, dot shape, evenness of
ink distribution (ink layer) are specification for the category “colour”. Sharpness,
addressability, gradation (tone value range) are attributes dedicated to the specification
“resolution”. Dot/colour separation position, and printed image position are attributes
attached to the specification “register”, while gloss, mottling and evenness belong to the
specification “surface”. Specific test elements (targets and patches to measure) are printed
59
together with the print image and appropriate measurements instruments (see Section 2.2)
are available for continous quality control. In addition, print quality can be checked
visually. However, some minimum requirements in terms of appropriate illumination and
viewing conditions according to ISO 3664 [79] must be fulfilled to perform a visual
quality control. Depending on the image quality factor definition there are a number of
other attributes which obviously can be considered as dimensions of image quality, such as
tone reproduction, dynamic range, contrast, noise, graininess, artifacts etc. Nussbaum et al.
[134] (Paper A) proposed a number of quality factors customized for the evaluation of the
appropriate digital print system.
Engeldrum [38], in his famous “Image Quality Circle”, is using the “Nesses”,
which reflects the ending of many words describing some the important perceptual
attributes. He is also using the expression psychometric scaling as the process in which
observers are asked to assign numbers to the perceptual attributes according to their visual
judgement. Typically, lightness, colourfulness, pleasantness, and sharpness are examples
of human perceptions reflecting these attributes. For further details on perceptual colour
attributes and their definition we refer to Hunt [70].
Very early two types of image/print quality assessment methods were recognised.
The first method is based on observation, using psychophysical experiments to gather the
judgement of human observers. The second method is by measurement, using instruments
to determine values for the various quality factors. These are mathematical models based
on perception.
2.4.3 Print Quality Evaluation by Visual Observation

The evaluation and judgement of print quality performed by human observers involves
psychophysical experiments. The discipline concerned with the measurement of perception
or sensation is psychophysics, which according to Fairchild [42], “is the scientific study of
the relationships between the physical measurements of stimuli and the sensations and
perceptions that those stimuli evoke.” And in the context of perceived image or print
quality methods have been developed to measure the human perception or sensation.
Historically, psychophysics has its origins in the early 19th century, when Ernst
Heinrich Weber was asking observers to lift a certain weight and then added increments to
the weight until the observers just realised the new weight different from the original.
Weber found that the ratio of ΔI/I is constant, whereby I represent the magnitude of the
stimulus (weight in this case) and ΔI is the change required in the stimulus (increment of
60
weights) to be just noticeable (the least difference that the observer still can perceive as a
difference). Another important work (built on Weber’s law) is the transformation of the
physical stimulus intensity scale into a perceptual magnitude proposed by Gustav Fechner.
Fechner’s law, which relates the physical magnitude to the perceptual magnitude by a
logarithmic function is built on two assumptions, firstly, that Weber’s law is valid and that
secondly a just-noticeable difference (JND) is a unit of the perception scale. This was
published in Elements of Psychophysics in 1860 [47].
Stevens studied the relationship between physical stimulus intensities and the
corresponding perceptual response for a large number of different types of perception. He
conducted experiments using magnitude estimation and magnitude production ratio scaling
to study brightness and loudness [55]. One of his particular improvements compared to
Fechner is the direct estimation of the perceived intensity using psychophysical ratio
scaling. It results in Steven’s power law S = οIm, where S indicates the subjective,
perceived intensity, I the physical stimulus intensity, ο is a constant (which depends on the
measurement units used) and m is the exponent which changes depending on the nature of
the physical stimuli judged (typically for brightness <1, more precisely 0.33).
2.4.3.1 Scaling types

In order to evaluate an image reproduction in terms of the subjective quality, a number of
methods can be applied to gather the observer’s judgements and to quantify the perceived
image quality considering the reproduction properties. Before the most common methods
are discussed four scaling types proposed by Stevens [175] have to be addressed. He called
them: nominal scale, ordinal scale, interval scale, and ratio scale.
Nominal scale is the simplest scaling type (also denoted as categorical), which uses
names or labels for certain characteristics to distinguish among them.
In an ordinal scale the assessed samples are assigned with numbers to represent the
rank order either in an ascending or descending order dependent on the judged attribute.
However, the distance between samples can be large or small and can change up or down
the scale. For instance a set of images with different intensity of sharpening applied could
by ranked from the most sharpen image to the least. Although, we can obtain a ranking
order from most sharp to least sharp or vice versa the scale does not provide information
about the magnitude of sharpness differences between the scale values, only information
about the greater than (>) or less than (<) property.
61
An interval scale adds the magnitude of distance between the images to the ordinal
scale. The differences between the scale values on the interval scales represent equal
distance anywhere along the scale. For example, if a pair of images (e.g. with different
sharpness) are judged and separated by two units and a second pair of images at some
other point on the scale represent the same two units of magnitude in difference, the
perceptual difference of the two pairs will be equal [39].
The ratio scale facilitates all the properties of the first three including a meaningful
zero-point. However, according to Engeldrum, defining a meaningful zero-point in visual
work determining image quality can be rather difficult [39]. For example, classifying a
grey image equal to zero colourfulness sounds straightforward. But, what about a hue scale
with a hue having a starting point of zero? Besides the mathematical operations addition
and subtraction, which can be performed with interval scaling, multiplication and division
are further operations, which can be used on a ratio scale.
Most of the applied psychophysical experiments are intended to perform on an
interval scale or a ratio scale, either directly or, by observer estimation or by data post
processing analysis according to an appropriate probabilistic model [103]. In general, it
can be concluded that the greater the power of the scaling type the more difficult the task
can be for the observer performing the appropriate judgement. In other words, performing
a nominal scaling by labelling or naming items or generating an ordinal scale ranking of a
set of images according to a “most sharp…least sharp” sequence can be considered as
rather easy tasks for observers. A much more demanding task for observers (at least
without training) is to respond with a certain number for the ratio to judge a given attribute.
2.4.3.2 Psychophysical methods

Essentially, four types of psychophysical methods, presenting the stimuli to the observers
and the technique of collecting the observer’s judgement have been proposed: Threshold,
matching, measuring differences and direct ratio scaling, see Figure 16. The first two
methods (threshold and matching) are dealing with discrimination each in its own way,
which means that threshold discrimination is aiming to find a stimulus that can be
distinguished as different from a certain standard stimulus. On the other hand, matching is
almost the opposite process: finding a stimulus that can’t be discriminated from a standard
stimulus. However, both methods can use the same discrimination technique.
62
For generating measuring difference for stimuli with supra-threshold perception,

four methods are frequently used: rank order, pair comparison, categorical judgement, and
rating. Their final results are given in an interval scale.
For obtaining a ratio scale, a direct ratio scaling method is used including
magnitude estimation, magnitude production, ratio estimation and ratio production.

Psychophysical methods

Measuring

Direct ratio
Threshold Matching

differences scaling
Method of Memory Rank order Magnitude

limits

matching

estimation

Pair
Method of Asymmetric comparison Magnitude
single matching production

stimulus
Categorical

judgement

Method of
adjustment
Graphical

rating

Figure 16: Overview psychophysical methods.
Threshold
Threshold experiments are designed to determine the JND. It is used to measure an
observer’s sensitivity to a given change of a stimulus. Engeldrum [38] describes thresholds
in two ways. Firstly, he gives a classical definition where the threshold is the amount of a
physical stimulus needed to evoke a JND. Moreover, thresholds are expressed as a physical
specification of the stimulus (e.g. a brightness threshold can be measured in luminance
units of cd/m2). His second view is based on his “Image Quality Circle” and the customer’s
perception on different “nesses”, which are typically lightness, colourfulness and
sharpness.
In general, threshold techniques are very useful for determining visual tolerances
such as those for perceived colour differences. The basic idea of JND or threshold is
related to the concepts of discriminal dispersion or the probability function. Observers are
typically asked “Do you see a difference?” or “Is stimulus A different from stimulus B?”
Although the observers responses might be simple “Yes” or “No”, large variations can
occur by cumulating the observers answers and the result is described by a probability
distribution, like the discriminal dispersion of Thurstone [181]. Three types of experiments
measuring thresholds have been evolved to determine absolute threshold and JND, which
63
are “methods of limits”, “methods of single stimulus”, and “methods of adjustments” [55].
Principally, they vary in the method presenting the stimuli, in recording the observer’s
response, and in the way data analysis is performed. For further details we refer to
Gescheider [55].
Matching
As mentioned previously, the matching method is rather similar to the threshold technique,
except that the aim is to determine when two stimuli are not perceptibly different.
Matching experiments provided the basis to derive the colour matching function of the CIE
standard. A colour match across some certain changes in viewing conditions are called
asymmetric matching. For example, a test stimulus viewed in daylight illumination has to
be matched to another stimulus viewed under incandescent illumination. The resulting pair
of corresponding colours can be used to determine and test colour appearance models
generated to compensate for different viewing conditions. For some experiments, observes
are asked to view two stimuli using each eye separately. Each stimulus is viewed under
different viewing conditions simultaneously and the observer’s task is to produce a colour
match. This type of matching technique is called haploscopic matching methods [42].
Another type of matching experiment is where observers are asked to produce a
match or to give a response to a previously memorized colour. For more details on
memory match under different viewing conditions see the work of Panak et al. [141].
Measuring differences
In the field of image reproduction the measuring differences method is very often used to
assess image quality and is intended to obtain a relationship between perceptual
magnitudes of a reference stimulus and a test stimulus. In such experiments observers are
asked to make their judgement on a single perceptible attribute or “ness” in terms of
“greater-than” or “smaller-than” properties of the viewed samples. Note, that for some
applications this kind of properties (ordinal scaling) might be sufficient. However, as
mentioned previously, to determine the magnitude between the samples (images) an
interval scale is a minimum requirement. Given an original and a reproduction (including a
set of different reproductions according to a certain “ness”), there are a number of different
methods of presenting the samples to the observers to obtain their judgements about the
various reproduction properties. The two most important psychophysical methods used in
the field of image reproduction are the rank order technique and pair comparison
64
technique. A third important method, which has to be considered for the evaluation of
image quality, is the category judgement and finally, the graphical rating method.
Rank order method

Asking observers to arrange or rank image samples in increasing or decreasing order, (e.g.
from best to worst) along a particular perceptual attribute, such as pleasantness, is a typical
task for this particular method as applied in the study by Nussbaum and Hardeberg [131]
(Paper D). A common data collection procedure is to ask the observer to assign the order
from 1 (best) to n (worst) reproduction, where n represents the number of reproductions to
be assessed. The obtained data set (averaged from J observers) is providing information on
an ordinal scale showing which reproduction is performing best, second best, etc.
However, the data does not express the magnitude of how much the reproductions differ.
In other words, although the reproduction ranked first and second could be very similar or
perhaps very different, it only possesses the greater-than property of an interval scale. On
the other hand, the rank order method is considered as rather simple for the observer to
perform. Furthermore, due to the simplicity of the method, the observer’s repeatability
performance is rather high, meaning that repeating the experiment at different times results
in similar answers.
Pair comparison method
In pair comparison, observers are asked to choose the stimulus (or reproduction) of a pair
that exhibits more of the desired reproduction property. For example, in the work
conducted by Nussbaum and Hardeberg [132] (Paper B) the observers were asked the
preference question in terms of “Choose the image of the pair you prefer in terms of
naturalness”. Suppose that a reproduction is judged to be the same (due to identical
appearance to the observer) in terms of the chosen property, the observer is either to be
forced to make a choice of the one or the other anyway (forced choice) or the observer is
allowed to point out a match. The proportion of times a reproduction is judged greater in
some particular attribute than another reproduction is calculated and recorded. Although
such pair-wise judgement results can immediately express a relative magnitude of
differences (due to the ordinal scale), they can’t give information about their absolute
variation. Hence, an interval scale from the pair comparison data has to be generated. A
further constraint of the pair comparison method (even though it provides the most
accurate results) is the large number of judgements required for all pair combinations. For
n stimuli the number of pair–wise combination is n(n-1)/2.
65
Thurstone’s Law of Comparative Judgement

As already discussed previously, when observers judge samples either with the rank order
method or the pair comparison technique they do not produce an interval scale to quantify
the difference between the stimuli. With the help of Thurstone’s law of comparative
judgement [181] the collected data can be transformed into interval scale data where z-
scores represent the distance of a certain stimulus to the mean, and compared to the other
stimuli being assessed.
Thurstone formalized a model for the judgement process, including various “cases”
of the model based on different assumptions made. His model is based on the notion that a
discriminal process (as he called the observer’s process of judging samples) performed by
observers produce variations in the response of judging a stimulus (due to fluctuations of
the organism) which was called the discriminal dispersion (nowadays it is named standard
deviation of responses). He postulates that the observer’s responses have a variable effect
on a hypothetical perception scale (psychological continuum), which follows a Gaussian or
a normal distribution.
Given two stimuli A and B being judged, the difference between them is determined by the
distance between the mean of their response distribution, which in general can be
calculated using the following formula:
ΨB − ΨA = Z AB σΨA
2
+ σΨB
2
− 2rΨAΨBσΨAσΨB (15)
The symbol Ψ is used to indicate the discriminal process generated by a particular

stimulus, where ΨA and ΨB represent the mean response distribution of the stimuli A and
B respectively, rΨAΨB is the correlation coefficient between the two distributions, indicating
the extent to which a sample stimulus from the distribution ΨB and a sample stimulus from
the distribution ΨA tend to be interdependent. σ represents the standard deviation for the
stimulus A and the stimulus B and z is the normal deviate or z-score corresponding to the
proportion of times stimulus A is judged greater than stimulus B according to some
perceptual attributes. Assuming having a stimulus A, which has been chosen over stimulus
B to a certain percentage, then the corresponding z-score is the distance from the zero
mean (on a scale where the unit represents the standard deviation of the distribution). The
distance is corresponding to the area under the normal distribution curve too, which again
is equal to the percentage of times that stimulus B was chosen over stimulus A. In other
words, z-scores have the effect of transforming the original normal distribution into a
66
metric which indicates how many standard deviations an observation is above or below the
mean. Figure 17 illustrates a normal distribution with zero mean and unit standard
deviation (i.e. the Gaussian or standard normal distribution). It can be seen that for an area
of 0.84 (i.e. 84 %) under the normal distribution, which represents the percentage of times
that stimulus B was chosen over stimulus A, the corresponding z-score is 1.00.
Consequently, this is also the distance between the two stimuli on the interval scale
obtained using the law of comparative judgement.
Normal
distribution Percentage=0.84
Probability
{
-4 -3 -2 -1 0 1 2 3 4
Standard deviation ( z-score=1.00
Figure 17:Normal distribution with given area percentage and corresponding z-score
(after Morovic [121]).
In practice, the values of the standard deviation σ and the correlation coefficient r are very
often unknown and can’t be measured experimentally. Therefore, Thurstone outlined six
cases (I-VI) applying different assumptions. Case V is most frequently used making the
assumption that σΨA = σΨB (i.e. discriminal dispersions or standard deviations are equal),
and that rΨAΨB = 0 (i.e. there is no correlation between the responses of different stimuli).
Given these assumption, case V can be simplified and computed as:
ΨB − ΨA = Z AB 2 (16)
For more details on cases I-IV and case VI see the work of Engeldrum [38], and
Gescheider [55].
The precision of the experimental results can be described in terms of the 95% confidence
interval (CI), which indicates the critical distances for significant differences. It can be
computed as:
67
σ
R ± 1.96 (17)
N
where R is the mean, σ reflects an estimate of the standard deviation and N represents the
number of observations. Because R has a scale of σ 2 , σ = 1/ 2 , thus the CI can be
calculated by:
( )
±1.96 * 1/ 2 / N (18)
Category judgement method

If the number of stimuli is large and the observer is asked to separate the stimuli into
various categories, the category judgement method is appropriate. This method involves
the definition of equally spaced categories in which the observer is asked to assign the
viewed sample to one of the defined categories, e.g. 1=No colour difference, 2=JND,
3=Noticeable colour difference, …, 7=Largest colour difference. This method requires
more judgement on the part of observers than pair comparison and is more suited to the
evaluation of a large number of stimuli which can be impracticable by the pair comparison
method.
The data analysis of the observer’s judgements is dependent on the assumption
whether the observations were made on a equally spaced scale (where the perceived colour
difference between e.g. categories 2 and 3 is equal to that between categories 6 and 7, etc.),
then the mean-category-value method can be applied or that the scale used by the
observers is not evenly spaced [13]. If that is the case a transformation of the observers
response has to be applied to obtain an even spacing by using Torgerson’s Law of
categorical judgement [183]. As discussed above, Thurstone formulated his Law of
Comparative judgement according to the scale difference between the samples to be
judged. However, Torgerson’s model, which is an extension to Thorstone’s model,
calculates the difference between the sample scale value and the category boundary.
Bk − ΨA = Z Ak σA2 + σk2 − 2rΨABkσΨAσk (19)
where Bk is the mean location of the Kth boundary, ΨA is the mean response to stimulus A,
σA is the discriminal dispersion (standard deviation) of stimulus A, σk is the discriminal
dispersion of the Kth boundary, rΨABk is the coefficient of the correlation between
momentary positions of positions of stimulus A and category boundary k on the scale.
68
Finally, Z Ak is the normal deviate corresponding to the proportion of times A is placed

below category boundary k.
By comparing Equation 15 and Equation 19 we observe the same form. However,
the difference between the two models is simply that Torgerson’s law of categorical
judgement relates to the relative positions of stimuli with respect to category boundaries
rather than with respect to another stimuli.
Graphical rating method

This method provides a direct interval scaling estimation. Here, the observers are asked to
indicate the magnitude of their perceptions on a one-dimensional scale that has defined end
points. For example, the observers are asked to judge the amount of sharpness in a set of
reproduction samples. The observer is given a card with a line of about 20 cm in length.
On both ends of the card are adjectives describing the property of the particular attribute
such as “Extreme sharp” on the left side and “Extreme un-sharp” on the right side. Now,
the observer is asked to indicate (by marking on the line) the physical distance on the line
as the distance between the viewed reproduction and the corresponding perceived attribute
scale. Consequently, the observer’s judgement provides an interval scale, which indicates
the mean location on the graphical scale for each stimulus. (Similar method can be applied
on the computer monitor recording the movement of the cursor distance.)
Direct Ratio Scaling Method

Previously, we have addressed methods of ordinal and interval scale generation using
different methods of data collections and analysis techniques. The following methods will
provide results in ratio scale applying the Steven’s law [176] which states the relationship
between a physical stimulus magnitude and its perceived intensity. Basically, there are two
types of ratio scale, magnitude estimation, and magnitude production, respectively.
Magnitude estimation
In the magnitude estimation the observer is asked to give a numerical response in
proportion to the perceived strengths of the particular attribute to be judged. Suppose an
observer is assigning the number ‘10’ to a darkness reference sample. Consequently, the
observer is asked to give a number that is twice the reference for a stimulus that is twice as
dark as the reference sample. The key point in the magnitude estimation is that the
numerical response of the observer should represent the ratio of the strength of the attribute
of the sample compared to the strength of the reference. This method has been applied in
69
the study by Nussbaum et al. [134] (Paper A) using an expert panel to conduct the
psychophysical experiment.
Magnitude production
The inverse of the magnitude estimation is the magnitude production. Essentially, the
observer is given a number and is asked to generate a stimulus that matches the perceived
magnitude compared to the given number. Due to difficulties of the appropriate
adjustments of the selected attributes this technique doesn’t have many applications in
imaging. However, it is assumed that with better understanding of image quality and its
component this technique will play a bigger role in the future [38].
2.4.3.3 Experimental set up

As we have seen previously, the aim of the scaling process is to have observers assign
numbers to certain chosen perceptual attributes. Independently of which scaling method is
used, the appropriate experimental setup is determining the human judgement. Therefore,
information regarding the viewing conditions, image selection, image size, type of
observer, and observer instructions has to be provided. The number of samples and
observers needed is task dependent. For example CIE 156 [30] “Guidelines for the
Evaluation of Gamut Mapping Algorithms” recommends to have at least 15 observers to
perform a psychophysical experiment using the methods of pair comparison, category
judgement, or ranking.
2.4.4 Quantitative Print Quality Evaluation

Applying print quality evaluation in the context of process control, the observer based
assessment method is a subjective process and is considered as time-consuming,
inconsistent, resource demanding and even expensive. A quantitative measure is usually
developed to take into account the human visual system and thus being correlated with the
subjective assessment.
As seen previously in Section 2.4.2 print quality can be affected by many factors
from the digital input to the print or from the print to the appearance. In case of a mismatch
between the original and the reproduction some of these factors may have contributed to
the distortion and the question can be asked which factors have caused the mismatch and
how can the distortion be assessed? Suppose the printing process is conducted according to
a certain conformance but the printed image is distorted in colour, then the digital input is
not appropriate transformed according to the given output parameters. On the other hand,
70
the digital input can be prepared for a certain output device but the printing conditions do
not conform to specification and do not match the input data. In the printing industry,
quality inspections are used to control process conditions, and to ensure that the end
products meet certain quality criteria. A common approach to measure print quality is to
make instrumental measurements of certain printed properties, such as colour, resolution,
register, and surface, and use a quantitative metric to determine the print quality. In this
work the property “colour” with its specifications colour accuracy, optical density, dot
gain, evenness of ink distribution (ink layer) in different printing workflows has been
determined by using quantitative colour measurement [130-132, 134, 172, 193, 194]
(Papers A-E). To determine the print quality in terms of process control the obtained
measurement data were compared to specifications (target values and tolerances) provided
by ISO [79, 80, 83-87]. Operating with quantitative evaluation using measurement
instruments the measurement uncertainty has to be considered, see Section 2.2.2.
As discussed in the previous section, colour management and ICC device profiles
are widely used in the graphic arts and printing industry to transfer colour information
between colour reproduction media. The print quality and the accuracy of a colour from a
given input, to the display, and further to the printed image in a colour management system
depends among other factors on the quality of the profiles involved. The assessment of
ICC profiles is a rather complex issue and involves a number of aspects to consider
depending on the users requirement and the colour image reproduction’s objectives.
Sharma [166, 168] proposed methods to evaluate the aspect of “colorimetric accuracy” of
device profiles considering in-gamut CIELAB values. As mentioned earlier, printer
profiles are typical two-way profiles; the invertibility assessment of an ICC printer profile
is considering the performance of the transformation from the PCS-to-device lookup table
(B2A LUT) and the device-to-PCS lookup table (A2b LUT). The "Round-Trip" test is a
useful method proposed by Sharma. The B2A LUT is relevant for printing an image while
A2B LUT is used when we preview or soft proof images on a monitor display or
performing a print simulation on a hard proof system. Depending on the chosen rendering
intent these transforms should be either accurate or pleasing. Green [58] provides a
recommendation to assist in the evaluation of the colorimetric and perceptual rendering
intent transform in ICC v4 profiles [72]. Another aspect in device profile evaluation is to
consider the performance in terms of “gamut mapping”. Different GMA’s are proposed (as
seen in the previous section) to transform colour values between gamut’s of different sizes.
The behaviour and performance can be evaluated by comparing GMA’s to each other.
71
Consequently gamut mapping affects the aspect “smoothness”, which is perceived when
transitioning between colour shades. In visual evaluations of colour reproductions the
aspect of smoothness is a very desirable property of colour transforms and often given a
high ranking. If the viewer perceives a lack of smoothness in a colour transformation it is
considered as an unexpected jump in the difference between adjacent regions [59]. A work
by Falkenstern et al. [44, 45] proposes a framework containing a number of aspects (e.g.
colorimetric accuracy, colourfulness, smoothness, gamut volume, etc.) and the
corresponding metric to quantify the performance of a ICC printer profile. It is known that
manufacturers of commercial profiling tools use different techniques to fit measurement
data into a model of the device to be characterized, thus each tool has its own strengths and
weaknesses. A study by Büring et al. [23] investigates profiling tools and the generation of
ICC profiles based on one single set of measurement data. The work focuses on image
quality asking observers to judge the reproductions of natural as well as artificial images
with various image content with respect to a given original.
It is worth mentioning other research areas in the field of quantitative print quality
evaluation. Considering print quality control Kuenzli et al. [108] proposed a mini-target
for on-line measurements in newspaper printing production using an RGB camera to detect
the target values and an image analysis software which is able to assess the registration,
solid-tone density, dot gain, and colorimetric values. A print quality toolkit which
calculates the print quality metrics specified in the ISO/IEC guidelines on office equipment
- measurement of image quality attributes for hardcopy output [96] is presented by Grice
and Allebach [61]. The printed test target is scanned at a desired resolution, and the
resulting scanned test target is analysed according to four distinct categories of printed
areas: line and character metrics, solid-fill (colorant blocks) metrics, tint solid (halftoned
blocks) metrics, and background field (non-printed areas) metrics. A similar approach,
using a scanner or camera to perform objective print quality measurement is presented by
Streckel et al. [180] and Kipman [104]. The authors present a series of specific metrics that
can be used for quantifying printed image quality for digital printer and paper
manufactures.
Recently studies have been presented proposing a framework for the evaluation of
colour prints using image quality metrics [144]. Multiple work by Pedersen et al. [145-
148] proposes quality attributes (such as colour, lightness, contrast, sharpness, and
artifacts) for the evaluation of colour prints. In combination with an appropriate image
72
quality metric these attributes are estimated and validated to be able to predict print
quality.
73
74
Summary of included Papers
3 Summary of included Papers

In order to demonstrate the relation between the papers and topics addressed, the presented
research area can be divided into categories and keywords or subjects. Three categories
can be pointed out. They are called “human”, “workflow” and “physics”, indicating the
type of core characterization. Moreover, each category includes different keywords
specifying the subjects. The category “workflow” is focusing on the subjects “print quality
evaluation”, “colour measurement”, “standardization”, “process control”,
“calibration/character-ization”, “standards/parameters” and “colour management”. The
category “human” considers the subjects “print quality factors” and “psychophysical
experiment”. The category “physics” is focusing on the subjects “measurement
uncertainty” and “measurement instrument characterization”, respectively. The bold letters
A, B, …, and G, as shown in Figure 18, represent the seven individual research papers and
indicate the main area and subjects of each contribution and how they relate to each other.
It is important to point out that the structure of categories and subjects is very dynamic in
the sense that most of the subjects have large intersections. The overlaps among the
subjects and between the categories indicate their relation.
Print qualityy
f
AB Psychophysical
ment
Human
D
Print quality evaluation
Colour measurement
Standardization Process
control
C Colour ma
Calibration /
characterization
Standards /
parameters
Workflow
E F Meas
unce
G
ment Physics
ent
zation
Figure 18: Diagram demonstrating the categories and subjects and the contribution of the
papers related to them.
75
Summary of included Papers
76
Summary of included Papers Paper A
3.1 Paper A
3.1.1 Abstract
Potential customers within the digital printer market have various demands considering
their requirements with regards to desired print quality. This paper aims to investigate the
print quality according to predefined quality factors to determine the appropriate printing
equipment. In particular the study describes the methods and results of a research project
conducted at the Norwegian Colour Research Laboratory for the Norwegian Government
Administration Service. The objective of the project was to develop methods and
procedures to perform test prints from digital printers including the evaluation and
assessment of the print quality according to predefined quality factors. The results of the
study indicate the performance of various digital printers in terms of their obtained print
quality. This will be useful for the governmental purchase decisions.
3.1.2 Motivation
After several years of market hesitation, digital presses have now become common, and in
today’s printing market, where flexibility, variable content, shorter lead times, and on
demand publishing, are being requested, digital presses represent an attractive supplement
to conventional offset presses [66, 105, 110]. The Norwegian Ministry for Transport and
Communications who is responsible for the design handbook defining the colours used in
the ministries logo (National Coat-of-Arms) has decided to embrace this new technology
and replace the conventional offset process for printing letterheads by digital printing
systems. However, such potential customers within the digital printer market have various
demands considering their requirements with regards to desired print quality.
In order to maintain the colour reproduction specifications, as outlined in the design
handbook, the Norwegian Colour Research Laboratory at Gjøvik University College in
collaboration with Norwegian Government Administration Service (NGAS) has carried out
a project to develop methods to evaluate an appropriate digital printer. Although a number
of studies and research have been done in the field of print quality it has repeatedly been
concluded to be a very complex issue. The subject has been discussed at various
conferences e.g. by Stokes [177], Hardeberg and Skarsbø [66], Yendrikhovskij [199] or
Norberg et al. [129].
77
Paper A Summary of included Papers
In the context of print quality evaluation this contribution [134] (Paper A) presents
a solution of developing methods and quality factors to evaluate digital printers in terms of
their potential impact on print quality.
3.1.3 Methods
Considering the methods applied to assess the digital print quality in this study we have
proposed a number of quality factors including their weights. Furthermore, the quality
factors have been divided into three assessment categories, namely, “visual logo
assessments”, “general print quality assessments” and finally the “copy quality
assessments”.
Principally there are two different methods when assessing print quality. The first
method is typically done by quantitative analysis, using measurement instruments to
determine values for the various quality factors. Considering the evaluation of relative
colorimetric colour reproduction of printing devices Microsoft has proposed a quality
assurance system [118]. In this study the quality factors including text quality, colour
gamut, repeatability and register are determined by quantitative analysis.
The second method is based on observation, using psychophysical experiments to
gather the judgement of human observers. In this work the magnitude estimation method
[55] was employed to judge the quality factors determining the colour logo reproduction in
terms of colour match, visual resolution, surface texture, logo alignment and artefacts. The
results are provided in ratio scale applying Steven’s law [176] which states the relationship
between a physical stimulus magnitude and its perceived intensity.
Four expert observers gave individual ratings on a scale from excellent (6) to very
poor (0), and the average results were listed by ranking. To complete the final ranking the
results of the three categories have been weighted once again according to the priorities of
the quality requirements defined by NGAS. The category “visual logo assessments” has
been weighted highest by 50% due to the importance of the colour match of the colour
logo reproduction, the category “general print quality assessments” by 30%, and finally the
third category “copy quality assessments” by 20%.
NGAS, as the customer of this project, nationally announced a printer quality
contest and printer manufactures were invited to participate in the test. Consequently
methods had to be designed, including the development of four different digital test
documents and test procedures. Finally, six competitors A, B,…, F have applied to conduct
the test.
78
Summary of included Papers Paper A
3.1.4 Result
The data analysis identifies competitor A as the candidate that scored best in all the quality
criteria in the category “visual logo assessments”. Especially for the quality factor “colour
match” which NGAS prioritised highest candidate A showed an excellent performance
followed by candidate B and C. Although the candidates D, E, and F have shown
reasonable results for the quality factors «surface texture», «register» and «artefacts» the
performance regarding «colour match» was rather poor. In fact, the expert panel has
decided that candidate D, E, and F will be excluded in the further evaluation process
because they couldn’t match the requirements of this category at all.
In the category “general print quality assessments” candidate B performs best
followed by the candidates C and A. Although candidate A has shown the largest range of
colours achievable compared to the other two competitors the repeatability performance
has shown some rather poor results. Finally, in the third category “copy quality
assessments” all three competitors achieved reasonable results without significant
differences among them. Lastly, although the final ranking doesn’t identify a significant
difference between competitors A and B in terms of weighted mean, competitor B has the
best score.
3.1.5 Conclusions
The evaluation method including the selection of the quality factors used in this particular
study has been unique in terms of print quality. Although various factors have been chosen
and applied in the evaluation process the main target to assess the competitors has been the
accuracy of the colour reproduction of the three colours of the national logo.
In this work we demonstrated and proposed categories and quality factors, which
can be applied and adjusted for other applications according to the customer’s
requirements to evaluate the appropriate print quality.
79
Paper A Summary of included Papers
80
Summary of included Papers Paper B
3.2 Paper B
Print Quality Evaluation and Applied Colour Management in Heat-set Web Offset
3.2.1 Abstract
This paper aims to investigate print quality in heat-set web offset by applying colour
management. In particular it looks at the colorimetric properties of five heat-set web offset
presses in order to evaluate the appropriate colour separation approach, either by applying
individual separation profiles or by using an industry standard profile such as
“ISOwebcoated.icc”. The key method relies on obtaining colour measurements to
determine the repeatability of each participant in terms of colour differences. Furthermore
the variation between the five heat-set web offset printing processes and the variation
according to the colorimetric values of the ICC profile “ISOwebcoated” are important
parts of the quantitative evaluation. According to the colour measurements two custom
ICC profiles were generated and applied to four test images, which were printed by five
heat-set web offset presses. Furthermore the industry standard ICC profile
“ISOwebcoated” was applied too. A psychophysical experiment was carried out to
determine naturalness of the reproductions made according to the three profiles applied.
Finally the results of the study indicate the performance of the appropriate profile applying
to the five heat-set web offset presses to obtain significant best print quality.
3.2.2 Motivation
Although process control for the production of half-tone colour separations, proof and
production prints are clearly defined in ISO 12647-2 [85] printing processes often show
major variations, which affect the appearance of print. The print variations can be detected
within the sheet, within the press or between the presses. Due to the print-on-demand
concept, different parts of a total publication edition can be printed in different print
locations.
However, the appearance of the total print edition must be identical. Similarly the
requirements for advertising campaigns’, which are published in different magazines but
printed on equal substrates, have to have identical appearance. Eventually, to strengthen
their position in the heat-set market, five of the largest Norwegian heat-set web offset
printing plants started a collaboration to evaluate their common print quality and print
control. Hence, the aim of this work [132] (Paper B) is to evaluate five heat-set web offset
81
Paper B Summary of included Papers
printing presses in terms of their conformance to specified values, in accordance with the
requirements of ISO 12647-2. Furthermore, the assessment of each individual printing
press and the variation within the five participants are important parts of this study, in
order to evaluate the appropriate colour separation approach, either by applying individual
separation profiles or by using industry standard profile such as ‘ISOwebcoated.icc’ [41].
3.2.3 Methods
To assess print quality a number of test prints produced under certain print conditions and
parameters are required. Often quantitative evaluation is used in combination with a
psychophysical experiment determining a certain quality attribute such as sharpness or
colour. In this work a colour measurement instrument has been used to investigate the print
run repeatability of each printing press, the variations between the presses and finally the
colorimetric variations in accordance with the target characterization table FOGRA28L.txt
[49].
The aim of the psychophysical experiment was to determine naturalness of the
image reproductions made according to different web coated prints. Four test images have
been prepared and converted according to one industry standard profile
‘ISOwebcoated.icc’ and two custom profiles. The comparisons between the printed images
were carried within each individual web offset press. The law of comparative judgement
was applied and the method used was pair comparison judgement [38]. The observer’s task
was to decide which of the two prints in the viewing cabinet was most preferred in terms of
naturalness.
3.2.4 Results
According to the colour measurement evaluation all printing plants show a short-term
repeatability (within press) performance within an acceptable tolerance. On the other hand
except for one printing plant the long-term repeatability shows variations, which can be
considered as rather large. In the course of our work, we have found that some of the
printing plants have changes in the calibration set up between the two test prints and
therefore the variations have been unveiled. An important factor to consider in terms of
print quality and print control is the variation performance between the presses. As a result
from the quantitative evaluation of the first test print run the variations between the five
printing plants demonstrate some undesired bias, which needs further adjustments. The
comparison between the obtained colour measurements and the target values of the
FOGRA28L.txt show colour differences, which again can be considered as rather large.
82
Summary of included Papers Paper B
Especially, for the primary colours solid magenta and solid yellow almost all printing
plants show colour differences which exceeds the defined ISO 12647-2 tolerances in both
test prints.
The results from the psychophysical experiment demonstrate that the industry
standard profile ‘ISOwebcoated’ performs significant best even though the calibration set
up aiming for ISO 12647-2 parameters is not fulfilled as seen in the quantitative
evaluation. Assuming that the print variations within the printing press between the first
and second test print would have been rather small it could be expected that the custom
profile (which is based on the first test print conditions) would perform best. However, as
seen from the long-term repeatability evaluation the print conditions in the second test
print have changed. Generally, industry standard profiles are related to a characterization
data set with a much larger number of measurements compared to the “custom” profile.
Hence, the industry standard profiles behave smoother in terms of preserving large print
tolerances than the “custom” profile. Moreover, the GCR (grey component replacement)
degree in the ‘ISOwebcoated’ industry standard profile is much stronger compared to the
custom profiles and therefore less sensitive to large areas with neutral colours. In other
words, the use of low degree of GCR can have a stronger impact to print variations and
finally affect the appearance of the print.
3.2.5 Conclusions
In this paper we have demonstrated a method to evaluate the print quality and print control
in heat-set web offset. Although the industry standard profile did perform best according to
the results of the psychophysical experiment some deficiencies have been detected. The
outcomes of the quantitative evaluation demonstrate clearly that there is still potential to
improve the target values of the ISO 12647-2 to obtain a better coherence between the five
heat-set web offset printing plants.
To be able to obtain the colorimetric target values defined in ISO 12647-2 [85] it is
very important to ensure the use of the appropriate ink set according to ISO 2846-1 [77].
This specifies the colour characteristics that have to be met by each ink in a process colour
ink set intended for heat-set web offset. Eventually the outcome in this study demonstrates
an obvious need for standardising the behaviour of the heat-set web offset presses.
83
Paper B Summary of included Papers
84
Summary of included Papers Paper C
3.3 Paper C
Implementing ISO 12646 Standards for soft Proofing in a Standardized Printing
Workflow according to PSO
3.3.1 Abstract
This paper defines one of many ways to set up a soft proofing workstation comprising a
monitor display and viewing booth in a printing workflow as per the Function 4
requirements of PSO certification. Soft proofing requirements defined by ISO 12646 are
explained and are implemented in this paper using a Nec SpectraView LCD2180WG LED
display along with Just colourCommunicator 2 viewing booth and X-rite EyeOne Pro
spectrophotometer. The display monitor colour gamut is checked for its ability to simulate
the ISO standard printer profile (ISOcoated_v2_300_eci.icc) as per the ISO 12646
requirements. Methods and procedures to perform ambient light measurements and
viewing booth measurements using EyeOne Pro spectrophotometer are explained. Adobe
Photoshop CS4 software is used to simulate the printer profile on to the monitor display,
while, Nec SpectraView Profiler software is used to calibrate and characterize the display
and also to perform ambient light and viewing booth measurements and adjustments.
3.3.2 Motivation
Related to print quality as seen in the previous two studies [132, 134] (Paper A and Paper
B) soft proofing has become a very important quality control check when simulating the
final print product. Although the concept of soft proofing is not new and the task may
sound rather simple, in practical applications the colour appearance between two different
media (e.g. soft copy simulation of a hard copy) can differ a lot due to unsuitable type of
devices, incorrect use of parameters and inaccurate device calibration and characterization,
or inappropriate measurement methods.
In the past a number of studies and research have addressed the issue of soft
proofing such as the work from Gatt et al. [53, 54], Leckner and Nordqvist [109], and
Roch et al. [156]. Katoh [102] discussed the appearance difference between the CRT soft
copy image and the hard copy image. Although ISO 3664 [78] defines the parameters and
tolerances for the appropriate viewing conditions and ISO 12646 [83] describes the
parameters and tolerances for the displays set up for colour proofing, the practical methods
to implement these standards as per the job requirements have not been clearly defined.
85
Paper C Summary of included Papers
Therefore the main contribution of this work [172] (Paper C) is aiming to describe in
details one of the ways to set up an appropriate soft proofing station (comprising a high
end monitor and a viewing booth with the appropriate ambient lighting conditions). The
proposed method is applied and verified according to ISO 12646 standard for soft copy
and hard copy proof comparison in the graphic arts industry. This is also in accordance
with soft proofing in a standardized printing workflow according to PSO [184].
3.3.3 Methods
The underlying experimental method is related to light and colour measurement using
commercial measurement instruments. Parameters for the ambient light setting, the
surround conditions, the viewing booth, the display calibration, and display
characterization are provided in the standards mentioned above. To measure and to verify
the set up of the soft proofing station including the ambient light conditions, commercial
measurement devices, used in the graphic arts industry were used. The obtained
colorimetric data were compared with the parameters provided in the standard.
Furthermore, the appropriate position and angle of the measurement instrument,
determining the measurements of the defined conditions, are demonstrated. The geometry
of the measurement device is very critical in terms of the measured values. Small changes
of the measurement instrument’s angle or the distance to the light source can affect the
measurement results dramatically.
3.3.4 Results
The measurement results of the monitor display analysis include the performance of the
white point and black point, grey balance, colour gamut, and the uniformity of the
luminance. The comparisons between the measured values and the target values in terms of
white point, black point and luminance show acceptable results. Although, the max ΔC*ab
is 2.35 units, the grey balance is still within the given tolerance. Flare light could be
considered influencing the measurement on dark colours. An important requirement of the
soft proof set up is the colour gamut of the display which should be as such, that it totally
encloses the colour gamut produced by the inks specified in the appropriate part of ISO
12647 for which the display is required to provide a proof [83]. According to the
measurement results the colour gamut of the monitor is big enough to simulate the ISO
coated printer profile. A further important quality criterion is the uniformity of the
luminance across the display. Different colour measurements (white, medium grey and
dark grey colour patches) are taken at nine positions on the monitor display. The obtained
86
Summary of included Papers Paper C
results demonstrate that the luminance is uniform across the display and corresponds to the
parameters defined in the ISO 12646 standard for soft proofing.
To analyse and evaluate the viewing booth and ambient lighting conditions,
measurements were performed using the X-rite EyeOne Pro spectrophotometer with the
ambient light adaptor. Subsequently, the measured values were compared against the P2
viewing conditions defined in ISO 3664 standards. The results show that the obtained
values are clearly within the given tolerances. Finally, implementation of the ISO standard
12646 alone does not guarantee that a displayed image (soft copy) will match the colour of
the same image produced on the hard copy in the viewing booth. Hence, to obtain an
appropriate colour match, colour transformation is required to convert the colour data from
the printer colour space to the display colour space by using a colour management system
and the appropriate ICC profiles [56, 58, 167].
3.3.5 Conclusions
In conclusion, in this paper we demonstrated a method to set up a soft proofing station
according to ISO 12646 standards for soft copy and hard copy proof comparison in the
graphic arts industry. The performance of the entire soft proof set up according to ISO
12646 has been evaluated. It was observed that, in spite of being within the standard’s
tolerance level the two images (soft copy image on the display and the corresponding hard
copy image in the viewing booth) might not show an exact visual match either. Therefore,
to investigate the disparity a psychophysical experiment could evaluate the magnitude of
visual differences between the soft proofed image and the hard copy in terms of
perceptibility and acceptability threshold.
Factors and their magnitude affecting the appearance on the monitor display and the
viewing booth need to be considered. New reflective monitor display technology, used in
combination with high level ambient light, needs to be investigated in terms of image
quality for soft proofing application. As a consequence, the colour measurement method
including technology and geometry needs to be verified.
87
Paper C Summary of included Papers
88
Summary of included Papers Paper D
3.4 Paper D
Print Quality Evaluation and Applied Colour Management in Coldset Offset
Newspaper Print
3.4.1 Abstract
This paper aims to investigate print quality in newspaper print by considering the
appropriate calibration standard and applying colour management. In particular, this paper
examines the colorimetric properties of eight Norwegian newspaper printing presses, in
order to evaluate the relevant colour separation approach, either by applying custom
separation profiles or by using an industry standard profile. The key method relies on
obtaining colour measurements to determine the repeatability of each participant.
Furthermore, the variation between the eight newspaper printing presses and the variation
according to the colorimetric values of the ISO 12647-3 standard are important parts of the
quantitative evaluation. Based on the colour measurements two custom ICC profiles were
generated and an industry standard profile “ISOnewspaper26v4.icc” was also used. The
first custom profile was generated using averaged colour measurement data set from a test
print run, and the second using a data set averaged between measured data and the
characterization data set “IFRA26.txt” provided by IFRA. These three profiles were
applied to four test images, which were then printed by the eight newspaper printing
presses. A psychophysical experiment was carried out to determine the ‘pleasantness’ of
the reproductions, which were produced using the three profiles. The results show the
performance of the appropriate profile, which is applied to the eight newspaper printing
presses to obtain significant best print quality. It reveals the importance of adopting
international standards and methods instead of using insufficiently defined house standards
to preserve equal results among different newspaper printing presses.
3.4.2 Motivation
Although process control parameters for the production of half-tone colour separations,
proofs and production prints, are clearly defined in ISO 12647-3 [86], often, newspaper
printing processes show significant variations which affect the appearance of print. The
Norwegian Newspaper Publishers’ Association (NAL) has provided three custom
newspapers ICC profiles in the period of 2000 to 2004 and recommends application of
89
Paper D Summary of included Papers
these profiles in the national newspapers printing process. However, these custom profiles
have two common characteristics in terms of their parameters.
Firstly, the number of colour measurements for generating the profiles is very
small, and secondly, the degree of GCR (grey component replacement) is rather low
considering newspaper print. The press calibration including the performance of the
profiles and the corresponding print quality has been considered as not satisfactory and
demonstrates the need for further revision. Therefore, the main contribution of this study
[131] (Paper D) is the evaluation of eight newspaper printing plants in terms of their
conformance to specified values in accordance with the requirements of ISO 12647-3.
Furthermore, the assessment of each individual printing press and the variation within the
8 participants are important parts of this study to evaluate the appropriate colour separation
approach, either by applying custom separation profiles or by using industry standard
profiles such as ISOnewspaper26v4.icc12.
3.4.3 Methods
Considering the print quality evaluation in this study, quantitative analyses based on colour
measurements and a psychophysical experiment has been applied.
The purpose of the quantitative evaluation is to determine the calibration process in
terms of the repeatability and uniformity performance [126] of each participant of the
project. Furthermore the variation between the eight newspaper printing processes and the
variation according to the colorimetric values of the ISO 12647-3 are important parts of the
quantitative analyses. This task requires a certain test document with IT8.7-3 CMYK target
including solid bars across the paper width to perform density control.
The colour measurement data measured on the IT8.7-3 CMYK Target from the
first Test Print were the starting point for generating a number of ICC profiles with
different separation settings applied to a number of test images. An expert panel
determined the two appropriate colour separations, which were then used to carry out the
psychophysical experiment using the test prints of the second test print run.
A total of two custom profiles and one industry standard ICC profile were applied
to each test image (by using the relative colorimetric rendering intent) and printed in all the
eight newspaper printing plants [58, 167]. The two custom profiles characterize the
12
IFRA, ISO Profiles. Available:
http://www.ifra.com/website/website.nsf/html/CONT_CONS_DL?OpenDocument&CTDL&E&
90
Summary of included Papers Paper D
specific Norwegian printing conditions, whereas the industry standard ICC profile
considers the conditions in accordance with the international standard ISO 12647-3.
The aim of the psychophysical experiment was to determine how pleasant the
reproduction of a newspaper print was considered to be, when compared to the remaining
newspaper reproduction prints. The observer’s task was to rank the three prints in the
viewing cabinet in order, from best to worst, in terms of preferred pleasantness [55, 63,
122].
3.4.4 Results
A calibration set up of a printing press including appropriate repeatability and uniformity
performance and the use of certain target values according to the given ISO standard
12647-3 are determined to obtain homogeneous print results among printing presses.
However, as seen from the quantitative evaluation results not only the variations in terms
repeatability and uniformity demonstrate some unwanted bias but also the colour
difference between the individual measurements of each printing press and the target
values show disparities, which partly exceed the defined tolerances. In particular the
results of the Tone Value Increase have shown large variations between the primary
colours cyan, magenta, yellow and black, between the printing presses and between the
two test prints. One of the factors, which have affected the differences, is the missing tone
value increase (TVI) specification in the Norwegian newspaper production method. Hence,
this can explain the inconsistency in terms of dot gain between the primary colours and
between the printing presses. Another possible factor affecting the print results is the
inappropriate measurement technology used in the newspaper production process
recommended by NADA which proposes to use image-based dot meters not only for
measuring the screening dots on the printing plates but also for measuring the dot gain on
newspaper substrates [128]. A study by Wroldsen et al. [194] (Paper E) investigated the
measurement performance of dot meters on newspaper substrate and concluded a very low
repeatability confidence using dot meters in newspaper print, compared to using a
colorimetric measurement approach.
The results from the psychophysical experiment demonstrate that one of the custom
profiles performs significant better than the second custom profile and the industry
standard ICC profile. However, it is observed that the measurement parameters of the
standard characterization data set do not match the calibration parameters of most of the
participants in Test Print 2. Hence, it is not expected that the standard ICC profile will
91
Paper D Summary of included Papers
perform better then the custom profile. Furthermore, the colorimetric difference between
the standard characterization data set and the custom characterization data set can be
classified as rather small. On the other hand the print variations within the 8 newspaper
printing plants in Test Print 2 are partly larger. Hence, the print inconsistency and not the
profile selection may have determined the visual print quality.
3.4.5 Conclusions
Although the quantitative evaluation has demonstrated some obvious shortcomings there is
a large potential for improving the target values of the ISO 12647-3 to obtain a better
coherence between the newspaper printing presses. Nevertheless, to preserve the daily
printing conditions and to match the colorimetric requirements of the adopted standard
profile it is highly recommended to perform press control according to a well defined
standard e.g. ISO 12647-3.
The main contribution of this paper is the demonstration of the importance of
adopting international standards and methods instead of using insufficiently defined house
standards to preserve equal results among different newspaper printing presses.
Furthermore the use of appropriate measurement instruments, combined with print
consistency in terms of repeatability is the fundamental requirements to obtain predictable
print quality.
92
Summary of included Papers Paper E
3.5 Paper E
A Comparison of Densitometric and Planimetric Measurement Techniques for
Newspaper Printing
3.5.1 Abstract
Two types of measurement technologies are used for process control in newspaper
printing, namely densitometric and planimetric technologies. Densitometric measurements
are done with densitometers or spectrophotometers, while planimetric measurements are
typically done with CCD image sensor-based instruments called dot meters. Although
these two technologies are fundamentally different, they are often used interchangeably in
print calibration and process control. In this paper we investigate the statistical relationship
between densitometric and planimetric measurements on newspaper print.
The aim of our project was to investigate whether it was possible to estimate
halftone values measured by a densitometer, from the halftone values measured by
different dot meters. The applied model is based on regression analysis using second order
polynomials. The results are given as estimates of the polynomial parameters, i.e. the
polynomials give the relation between halftone measurements with one of the dot meters
and halftone measurements with the densitometer.
Our statistical analysis showed that due to the large uncertainty of the estimated
parameters, the model does not accurately describe the relationship between the two
measurement technologies. This can be explained in part by the poor repeatability
performance for dot meters applied to newspaper print. Moreover the measurement results
also have shown significant variations within the three dot meters used in this experiment.
3.5.2 Motivation
For process control in newspaper printing, essentially two types of measurement
technologies are used, namely densitometric and planimetric measurements. In
densitometric measurements, the optical density is measured, and if needed converted to
halftone values, typically using the Murray-Davies equation [127, 182]. In planimetric
measurements, it is attempted to directly measure the halftone values, that is, the dot area
coverage, typically using dot meters containing a CCD imaging sensor [34]. Although
print calibration and process control, in particular in the Norwegian newspaper industry [1,
93
Paper E Summary of included Paper
128]. This motivated us to investigate whether there is a statistically significant

relationship between halftone measurements on newspaper print done with densitometers
(converted into halftone value with the Murray-Davies-equation) and halftone
measurements done with dot meters. Eventually the objective of this study [194] (Paper
E) is to find out whether it is possible to convert planimetric halftone measurements into
densitometric halftone measurements and vice versa. Since these technologies are used
interchangeably, it is important to know how to convert from planimetric measurements
into densitometric measurements, to keep the printing process under control and to achieve
predictable print quality.
3.5.3 Methods
To obtain measurement data with the measurement devices (one densitometer and three
dot meters), a specific test target consisting of 16 patches in different halftone values for
each process colour (CMYK), was designed and printed in coldset web-offset lithography
using AM-screening, in three different newspaper printing plants. Given the halftone
values measured by one of the dot meters, the aim was to predict halftone values of the
densitometer based on regression analysis using second order polynomials (Third order
polynomials tend to result in over-fitting). The results are given as estimates of the
polynomial parameters, which means that the polynomials give the relation between
halftone measurements with one of the dot meters and halftone measurements with the
densitometer. Furthermore, the measurement data were divided into two sets; training set
to establish the model (one model for each process colour) and a test set to evaluate its
performance. The residuals between the predicted and measured halftone values with the
densitometer (with test set = training-set and test set ≠ training set, respectively) were used
to judge the performance of the model. Because of significant measurement differences
between the dot meters and between the process colours for each densitometer – dot meter
– combination, it was necessary to study both the instrument combinations and the process
colours individually.
To justify that the densitometer could be used as a reliable representative for all
densitometers, we did a preliminary test with two different densitometers to verify whether
different instruments give the same result. The variation was less than 0.01 density for all
the process colours which is within the tolerance according to DIN 16536-2 [36]. A
repeatability analysis was also conducted for the three dot meters measuring on newspaper
94
Summary of included Papers Paper E
print the 50%-patch for each process colour (CMYK) 10 times. Consequently, the
repeatability results show rather large variations for all involved dot meters.
3.5.4 Results
Due to the fact that the correlation coefficients were close to 1, it can be assumed that there
must be a correlation between the halftone values measured by the various instruments.
Based on three measurement series of one test target, second order polynomials were
established estimating the relationship between halftone measurements with the dot meters,
and the corresponding halftone measurement with the densitometer.
If the test set is part of the training set, it is expected that the differences between
the predicted and measured halftone values are rather small, as the results have shown.
Furthermore, according to the results the variations are colour and halftone value
independent and do not follow a certain trend. However, there is no significant trend for
the obtained variations. Although the model performs well it is important to test the model
with another test-set.
When the test-set is not part of training-set the model does not perform that well
and the differences between the predicted and measured halftone values are larger.
Although the residuals have increased, the variation still does not follow a certain trend.
Similar results are obtained for the two other combinations of instruments. The low
repeatability of the dot meters is an unfavourable factor that makes our model
unsatisfactory.
Various factors affecting the repeatability and determining the performance of the
model are discussed in the paper. Print irregularities cause noticeable differences in
measured halftone values and reduce the repeatability when the aperture is small (in
combination with large halftone dots used in newspapers. Due to high optical dot gain
(especially for the middle tones and in newspaper print) it is ambiguous to decide what is
substratum and what is part of a screen dot. Large residuals between predicted and
measured halftone values for the middle tones could partly be explained by the high optical
dot gain and problems due to determination of threshold (what is substratum and what is
not) in the image analysis.
3.5.5 Conclusions
As expected the residuals between predicted and measured half tone values with the
densitometer increased when the test set was different from the training set. Moreover, the
measurement results have shown significant variations within the three dot meters.
95
Paper E Summary of included Paper
Overall, the general conclusion that can be drawn is that the model does not
accurately describe the relation between the two measurement technologies due to large
uncertainty of the estimated parameters. This can be explained by the poor repeatability
performance for dot meters applied in newspaper print. Eventually the main contribution of
this paper is to demonstrate that dot meters are not recommended for halftone
measurements on paper substrates in newspaper printing. Dot meters are originally
developed for measuring printing plates only.
96
Summary of included Papers Paper F
3.6 Paper F
Analysis of Colour Measurement Uncertainty in a Colour Managed Printing
Workflow
3.6.1 Abstract
Since the recent revision of ISO 12647-2 and ISO 12647-7, specifying the requirements for
systems that are used to produce hard-copy digital proof prints, the use of colour
measurement instruments is even more than before required in the printing industry.
Therefore, in a modern colour managed printing workflow, most of the printing houses use
more than one colour measurement instrument, typically one instrument in each
department (pre-press, press, and post-press). In this paper, a total of nine commercial
spectrophotometers are compared in terms of measurement uncertainty, precision and
accuracy, repeatability and reproducibility. The BCRA series 2 ceramic gloss tiles are used
to confirm the accuracy and repeatability of these measuring instruments according to the
manufacturer’s standards. We focus especially on inter-instrument and inter-model
reproducibility and discuss the effect of instrument calibration and certification.
For our experimental setup, four different materials are used, one proof print, one
commercial print, and one reference print, along with the BCRA series 2 ceramic gloss
tiles. In a colour managed printing workflow the use of more than one instrument can
impair and complicate the colour process control due to the colour differences between
different measurement devices. The effect of the colorimetric measurement errors due to
large inter-instrument and inter-model variability between instruments used in different
parts of the workflow is discussed and demonstrated.
3.6.2 Motivation
According to the findings in Paper B, D and E measurement instrument uncertainties can
highly affect the print quality in a colour managed printing workflow. Especially, the use
of more than one instrument in a printing workflow (pre-press, press) can generate
unwanted bias in terms of measurement disparities. One measurement instrument may
result in measurements which will qualify the print result to pass, and the other one can
result as a false-positive.
Various studies and research considering measurement uncertainties and
comparative studies of different types of colour measurement instrumentations have been
97
Paper F Summary of included Papers
presented in the past [18, 19, 137, 149, 157, 196, 197]. In a work by Ohno [138]
uncertainty is defined as an estimate of the range of values within which the true value lies.
However, our main contribution in this article [136] (Paper F) is the analysis of colour
measurement variability of using a number of different bidirectional colour measurement
instruments in a colour managed printing workflow.
Therefore, the aim is to evaluate the performance of nine commercial
spectrophotometers (one bench-top and 8 hand-held) typically used in the graphic arts
industry in terms of precision (repeatability and reproducibility) and accuracy.
Furthermore, the effect of colour measurement variability in a colour managed printing
workflow will be demonstrated. In particular the results of inter-model agreement
measuring colours on paper substrate will be reviewed.
3.6.3 Methods
ASTM E2214 [9] defines the specification, methods, and procedures evaluating the
performance of colour measuring instruments in terms of repeatability and accuracy.
Furthermore the reproducibility [197], which determines the variations between
instrument’s readings can have a huge impact on the evaluation of print quality in the
process control. Hence, it is an important aspect of this study. 14 BCRA series 2 ceramic
gloss tiles13 and their “true values” are used to confirm the accuracy of the instruments.
The repeatability performance of the measurement instruments is compared according to
the instrument manufacturer’s standards. We focus especially on inter-instrument and
inter-model reproducibility and discuss the effect of instrument calibration and
certification.
In order to analyse the measurement uncertainties of the colour measurement
instruments on commercial printed substrates, measurements were conducted on three
different paper substrates measuring the UGRA/FOGRA Media Wedge CMYK [162]
including 46 colour patches.
All instruments used in this paper reported spectral reflectance factor values from
380nm to 730nm with 10nm interval. Subsequently, the data were converted to CIEXYZ
tristimulus values according to the CIE 1931 2° observer and the CIE Standard illuminant
D50 using the method proposed by ASTM 308, Table 1 [5]. Furthermore, CIELAB (D50
as the reference white) values were calculated according to CIE 15 specifications.
13
BCRA, tiles are produced by CERAM Research, Queens road, Penkhull, Stoke-on-Trend, ST4 7LQ,
England
98
Summary of included Papers Paper F
Colorimetric difference ΔE*ab and ΔE*94 (as some manufacturers quoted the colour
difference in ΔE*94) values were computed between the BCRA reference data and the
measurements data obtained using each instrument [25].
3.6.4 Results
As stated in ASTM E2214 the most important specification in colour-measuring
instrument is repeatability. According to our results, except of two instruments, all others
did pass the manufacturer’s agreement in the short-term repeatability test and all in the test
available instruments did conform to the long-term agreement. The obtained results from
the inter-instrument test have demonstrated a reasonable performance among instruments
within the same family, as agreed by the manufactures. Nonetheless, the inter-model
measurement results have shown much larger colour differences, especially using
instruments from different manufactures. The results of the print measurements did
confirm the inter-instrument and inter-model agreement as observed by measuring the
BCRA tiles.
Considering the accuracy it can be observed that the chromatic BCRA tiles (e.g.
Red, Orange and Yellow) produce larger colour differences than the achromatic BCRA
tiles, perhaps due to possible thermochromism [43]. Furthermore, all the instruments
demonstrate smaller colour differences for the Black tile than for the White tile. The dark
to light grey tiles show a very similar behaviour. The larger differences on the White tile
may indicate that the instruments do not agree well on the definition of white, which could
be traced to the instruments’ calibration or certification status. However, there is no
obvious consistency observed between certified and non-certified instruments in terms of
their performance on the White tile.
3.6.5 Conclusions
Measurement instruments with valid certification and instruments with expired
certification have been used in this study to be conforming to the common situation in the
printing workflow. I might be expected that instruments with valid certification perform
better than instrument with expired certification. Although missing instrument re-
certification did not show a significant effect on the measurements in terms of reduced
accuracy or reproducibility performance, it is highly recommended to maintain the
instruments according to the manufactures requirement including appropriate re-
certification procedures. In conclusion, when applying only calibrated and certified
instruments, a further obvious consequence will be the use of only one certain instrument
99
Paper F Summary of included Papers
family (same model, same design of instrument from the same manufacturer with the same
parameters) in a colour managed printing workflow. This will minimize the colour
differences. To avoid large measurement errors among different instrument within the
same workflow it is required to use common filter specifications. In order to improve the
colorimetric performance and inter-instrument and inter-model agreement a method of
characterizing measurement instruments using colorimetric regression technique can be
considered.
100
Summary of included Papers Paper G
3.7 Paper G
Regression based Characterization of Colour Measurement Instruments in Printing
Applications
3.7.1 Abstract
In the context of print quality and process control colorimetric parameters and tolerance
values are clearly defined. Calibration procedures are well defined for colour measurement
instruments in printing workflows. Still, using more than one colour measurement
instrument measuring the same colour wedge can produce clearly different results due to
random and systematic errors of the instruments.
The aim of this paper was to determine an appropriate model to characterize colour
measurement instruments for printing applications in order to improve the colorimetric
performance and hence the inter-instrument agreement. The method proposed is derived
from colour image acquisition device characterization methods, which have been applied
by performing polynomial regression with a least square technique. Six commercial colour
measurement instruments were used for measuring colour patches of a control colour
wedge on three different types of paper substrates. The characterization functions were
derived using least square polynomial regression, based on the training set of 14 BCRA
tiles colorimetric reference values and the corresponding colorimetric measurements
obtained by the measurement instruments. The derived functions were then used to correct
the colorimetric values of test sets consisting of 46 measurements of the colour control
wedge patches. The corrected measurement results obtained from the applied regression
model was then used as the starting point with which the corrected measurements from
other instruments were compared. The goal is to find the most appropriate polynomial,
which results in the least colour difference. The obtained results demonstrate that the
proposed regression method works remarkably well with a range of different colour
measurement instruments used on three types of substrates.
3.7.2 Motivation
As seen previously, using measurement instruments in a print quality control context
reveals questions about the reliability of the obtained measurements. Especially when a
certain target colour is measured with different measurement instruments in different
locations and the results are qualifying the print differently. According to the results
101
Paper G Summary of included Papers
presented in Paper F [136] an obvious consequence will be the use of only one certain
instrument family in a colour managed printing workflow to preserve reasonable colour
differences. However, firstly this is practically almost impossible to implement and
secondly even within the same instrument family the measurements will vary. Eventually,
in order to improve the colorimetric performance and inter-instrument and inter-model
agreement, a method of characterizing measurement instruments has to be considered.
Therefore, the main contribution of this study [133] (Paper G) is to propose a
method to reduce the colour difference of colour measurement performed with more than
one colour measurement instrument measuring the same colour target. In particular, the
main contribution of this study is in correcting the measurements obtained by instruments,
using colorimetric regression technique. Finally, the appropriate correction model applied
to the measurement data sets will reduce the colour differences between the measurements
acquired by a master measurement instrument and the measurements performed by a
second measurement instrument used. Eventually, the model will improve the colorimetric
performance and inter-instrument and inter-model agreement.
3.7.3 Methods
The method we propose in this work is based on colour image acquisition device
characterization, which has been applied by implementing polynomial regression with least
square technique. Polynomial device characterization technique with least squares fitting
for different application has been adequately applied in a number of other studies by Kang
[101], Sharma [169], Hong et al. [69] and Johnson [100]. Compare to other studies where
the authors, e.g. Berns and Petersen [17] propose the correction of various systematic
errors by applying the spectral measurements data using multiple linear regression based
on modelling the results to improve the colorimetric performance, in this work, the aim is
to correct the instrument’s systematic errors by applying regression technique direct to the
measured CIELAB data.
Generally, to create a characterization model, a training set, existing of a reference
data set and the corresponding measurements performed by a measurement instrument is
required. In this work the model (training set) is derived by the 14 BCRA reference data
set and the corresponding measurements obtained by the measurement instrument. The
idea of using the BCRA tiles for the training set is to make the appropriate adjustments in
terms of the accuracy. Note, not only the precision between the instruments has to be
improved but also the accuracy in terms of the appropriate colorimetric values. Therefore,
102
Summary of included Papers Paper G
the BCRA tile reference values, which are traceable, have been used to establish the
model.
Eventually, the model has been tested with independent data using 46 patches of
the UGRA/FOGRA Media Wedge CMYK [162] on three different types of printed
substrates namely a hard-copy digital proof print, paper substrate type 1 and paper
substrate type 5.
Six commercial industrial-oriented spectrophotometer measurement instruments
typical utilized for daily production control in prepress and press applications are used.
One particular measurement instrument has been defined as a reference (“master
instrument”). The reference is not meant to represent the best or ideal instrument but
instead to be a state, to which to compare the measurements obtained by other instruments.
The corrected measurement results obtained from the regression model was then used as
the starting point with which the corrected measurements from the other instruments were
compared and eventually to find the most appropriate polynomial, which results in the
least colour difference.
3.7.4 Results
Even though the training set in the first attempt was limited to 14 samples only due to the
available BCRA tiles, the model performed excellent. The colour differences between the
corrected measurements of two measurement instruments could be reduced significantly
for all three types of substrates. In particular the instrument combination from different
instrument families demonstrate a noteworthy performance. Although there is no particular
function, which performs best for instrument combination within the same product family,
most of the polynomials reduce the colour difference with almost more than half compare
to the uncorrected data. Regardless the very small colour differences of the uncorrected
measurement data between inter-instruments (the same model from the same manufacturer
using equal specifications), applying the regression model have further minimized the
colour differences.
By extending the training set from 14 samples to 38 samples including the 24
colours from the ColorChecker the results does not significant improve in terms of
reducing the mean colour difference for all the measurement. However, the maximum
colour difference between all instrument’s combinations could be reduced substantially on
all three substrates. Hence, this indicates clearly that the model with 38 sample points is
more robust.
103
Paper G Summary of included Papers
3.7.5 Conclusions
The paper concludes that first order polynomials (more precise 3x5 polynomial) produces,
in most of the cases, the best results in terms of reducing the colour differences between
the instruments on different substrates. Although, there is no significant difference in the
performance of the model on the three different types of substrates, the proofing substrate
results in the least colour differences. Moreover, with instruments used from different
product families the inter-model agreement can be significantly improved, by reducing the
colour variations between the measurement instruments with more than half by applying
the characterization model. Increasing the size of the training set from 14 to 38 samples is
slightly reducing the maximum colour differences, but much more important, the model’s
behaviour is more robust in terms of different applied polynomials. To justify whether
thermochromism [43] affected the model, the BCRA tiles red, orange and yellow could be
left out in the training sample.
The main contribution of this paper is the demonstration of a method to correct the
instrument’s systematic errors by applying regression technique direct onto the
measurement output values in the CIELAB colour space. As a result, the colorimetric
performance and hence the inter-instrument and inter-model agreement has improved. In
conclusion the proposed method has shown significant improvement for taken
measurements with different instruments on reflective light. Therefore, it can be expected
that the method perform well too, on measurements obtained from light emitting devices
(e.g. display), including the use of different models of spectrophotometers,
spectrocolorimeters and colorimeters. However, this assumption has to be proofed.
According to ASTM E2214 [9] the most important specification is the
repeatability, including the reproducibility. Taking that into account an alternative
approach can be considered establishing the model using samples with similar or equal
physical properties for both the training set and the test set. In the current work we have
trained on 14 BCRA tiles to make the appropriate adjustments in terms of the accuracy and
tested on different paper substrates; which means the physical properties between the
training and the test set is different. However, a study by Steder et al. [174] illustrates that
the inter-instrument agreement increases by using samples of equal physical properties for
both the training and test set. In our work the model could be trained on printed sample
colours and tested on another set of printed colours on the identical paper substrate.
104
Discussion and Conclusion Discussion of Papers in Context
4 Discussion and Conclusion

After each individual paper has been summarized in the previous chapter the discussion
and conclusion of the thesis will be presented. In the first section the contribution of the
papers will be discussed in the context of demonstrating the relation between them.
Consequently a summary of the thesis contribution will be presented, before a conclusion
can be drawn und some ideas for future work are proposed.
4.1 Discussion of Papers in Context

The goal of this thesis has been to develop methodologies and procedures for print quality
assessment and process control according to objective evaluation and perceptive print
quality judgement. Because the quantitative evaluation has been an important part of the
study, the investigation of instrument measurement performance in terms of accuracy and
reproducibility for process control and print quality assessment was a focal point of the
study. Hence, a method was proposed to reduce the colour differences of colour
measurement performed and simultaneously improve the colorimetric performance and
inter-instrument and inter-model agreement.
Predictable colour reproduction has always been an important issue and a demand
within the graphic arts and printing industry. Although common standard parameters to
calibrate and characterize devices in the printing workflow have been defined and provided
in recent years or even decades, the implementation and the maintenance of the parameters
was not always prioritized by many companies. About the motive can be speculated.
However, it can be assumed that either missing competence or low market competition are
possible reasons for this absent priorities.
In the context of print quality assessment the seven papers associated to different
categories have contributed to the final present thesis. As illustrated in Figure 18 the
contribution of the papers generally cover the three categories “human”, “workflow” and
“physics” which encompasses the main subjects “colour measurement”, “print quality
evaluation” and “colour management” in this work. Even though Papers A [134], B [132]
and D [131] are mainly classified in the “human” category, because the subjects “print
quality factors” and “psychophysical experiment” are involved, the main contribution can
be related to the category “workflow”. Paper C [172] in the category “workflow” is
105
Discussion of Papers in Context Discussion and Conclusion
mainly dedicated to addressing the subject “colour management” in the sense of proposing
a method of applying ISO 12646 standards for soft proofing in a standardized printing
workflow according to PSO. In the course of our research work, we have recognized the
importance of colour measurement and the consideration of measurement uncertainties in
general. Therefore a very important contribution in the category “physics” are the subjects
“measurement uncertainty” and “measurement instrument characterization” represented by
the Paper contributions E [194], F [136], and G [133].
Recently offset and newspaper printing have adopted uniform standards including
parameters and tolerances and process control systems developed by a number of research
and standard groups such as ISO (International Standardization Organization), aiming to
stabilize and monitor the production processes [143]. It has been argued that the type of
standards used in production (official international or in-house standards) is not as
important as the fact that some form of standardization regulates the production process to
ensure that the same conditions are applied for each sub-process, e.g. from plate output
through to actual printing process. However, in our research we experienced great
misunderstanding and confusion in applying standards and parameters in the printing
industry. Among collaborating printing partners we have noticed the absence of quality
and tolerance agreements. A study by Martin et al. [117] confirmed our assumption that a
successful implementation of a colour managed printing workflow is rather difficult and
that achieving printed results that satisfy the customer is often a complicated business
involving a number of print-proof-adjust-reprint cycles.
In the beginning of the current research process there was no obvious distinction
between colour management and standardization of printing processes. During the research
period the difference became more noticeable in the sense that a standardized process is all
about calibration, including the aim of obtaining certain target print parameters, and the
preservation of the defined tolerances. Consequently, predefined industry standard profiles
corresponding to the applied calibration parameters can be used. On the other hand colour
management is dealing more with the optimization of a printing process without
considerations of any external specifications or standards, and therefore the need for
generating custom ICC profiles occurs. Nowadays, it can be argued that a standardization
process in printing is all about aiming at certain target values, performing quantitative
evaluation and implementing procedures to ensure that the target values are within certain
defined specifications and tolerances. Hence, conducting psychophysical experiments
might be redundant in this context as discussed in Paper B [132] and Paper D [131].
106
Discussion and Conclusion Discussion of Papers in Context
However, the implementation of standardization procedures is more then just applying and
maintaining certain process parameters. For a successful implementation high competence
and deep understanding of the entire colour managed printing workflow is needed among
the people who are responsible for colour management and colour measurements.
Therefore, conducting psychophysical experiments involving the production operators by
introducing standardization procedures is psychologically and strategically important and
is supporting the awareness of the standardization benefits. Together with the use of
quantitative instrumental measurement technique this will strengthen the prediction of the
expected print quality.
Figure 19 demonstrates that multiple aspects and factors contribute to the final
appearance of the print and hence affect the print quality. In this work, we have
systematized the various aspects interacting with each other. To determine print quality,
the definition of print quality factors and the different kind of methods applied to assess
them are as important as the concept of process calibration and standardization including
colour measurement for process control. For example in Paper A [134] we proposed a
number of quality factors to determine the colour reproduction performance of digital
printers.
As discussed in Section 2.4 the aspect of print quality can be determined either by
the image, its content and its transformation or by the behaviour of the reproduction
medium itself and the corresponding specification conformance applied. In the context of
standardization, standards and parameters, the evaluation of the process and the
reproduction media involved are important to determine weather the device performance
conforms to specifications. Hence, applying the appropriate calibration parameters and the
equivalent characterization is fundamental for the appropriate colour transformation.
Quantitative evaluation is based on using measurement instruments to determine
process control and print quality. The term process control in this work refers to the quality
assurance and prediction in terms of maintaining the print production within predefined
tolerance specifications. Since the recent revision of ISO 12647-2 and ISO 12647-7
specifying the requirements for process control for the production of half-tone colour
separations system and proofing processes working directly from digital data the use of
colour measurement instruments is more than before required in the printing industry. In
addition ISO 12646 which specifies the minimum requirements for the properties of
displays to be used for soft proofing of colour images together with our proposed method
implementing the standard presented in Paper C [172] the need for objective measurement
107
Discussion of Papers in Context Discussion and Conclusion
instruments is obviously. The increasing use of different measurement instruments in the

colour reproduction workflow, and as a result of that, the related uncertainty affecting the
reliability of the measurement data obtained has to be considered. As discussed in Paper F
[136], and Paper G [133] not only calibrated and certified measurement instruments
should be used but also the appropriate instrument type has to be applied for certain
measurement tasks. The outcome of the research in Paper E [194] clearly demonstrates
that dot meters are developed for measuring printing plates only and are not recommended
for halftone measurements on paper substrates in newspaper printing. To minimize the
colour differences between different types of instruments the use of only one particular
instrument family (same model, same design of instrument from the same manufacturer
with the same parameters) is recommended. Furthermore, in order to improve the
colorimetric performance and inter-instrument and inter-model agreement we proposed a
method of characterizing measurement instruments using a colorimetric regression
technique.
Print quality factors Standardization

Standards and parameters
Assessment methods
Process control
Print
quality
Colour management
Colour measurement
Concept of calibration, characterization

and colour transformation
Figure 19: Aspects and factors affecting and contributing to the print quality.
The methodology and procedures obtained during our research and the contribution of the
individual papers could be used as an application independent framework for those
involved in the process of print quality assessment and Process-Standard Offset (PSO). We
believe that the media community will benefit from the results and contribution of the
present work in terms of producing predictable print quality by applying the procedures
and methods proposed. Thus, we hope that we have - at least partially - reached the goal
that we aimed for at the outset of this project.
108
Discussion and Conclusion Summary of Contribution
4.2 Summary of Contributions

This section summarises the scope and contributions of the thesis. The basic objective of
this thesis was to propose methods and procedures to evaluate print quality in terms of
process control obtaining predictable colour reproduction. An important part of the work is
the investigation of measurement devices in the graphic arts and printing industry in terms
of their accuracy and reproducibility for process control and print quality assessment.
Thus, the measurement uncertainties were analysed and a method proposed to reduce the
colour differences of colour measurement performed with different instruments. The
proposed model improves the colorimetric performance and inter-instrument and inter-
model agreement. The substantial and innovative contributions of the research are as
follows:
• We developed methods and quality factors to evaluate digital printers in terms of

their potential impact on print quality. The proposed framework including
categories and quality attributes can be applied and adjusted for other applications
according to the customer’s requirements to evaluate the appropriate print quality.
The development of digital test documents and test procedures to obtain the
evaluation results is an important part of the concept.
• With respect to evaluation of print quality and print control in heat-set web offset
we have demonstrated a method applying quantitative assessment and in
combination with a psychophysical experiment determining the quality attribute of
naturalness. The presented method demonstrates the significance of applying
common standard target parameters, so that the heat-set web offset printing process
is consistent among and between printing presses to provide predictable print
quality.
• Even though ISO standard 12646 defines parameters for monitor and viewing
booth condition setup for soft proofing environment a clearly defined method to set
up an appropriate soft proofing workstation is not stated. We proposed a method
and procedure to set up a soft proofing station according to ISO 12646 standards
for soft copy and hard copy proof comparison in the graphic arts industry and
evaluate the performance of the entire soft proof set up according to ISO 12646.
109
Summary of Contribution Discussion and Conclusion
• We demonstrate the importance of adopting international standards and methods

instead of using insufficiently defined house standards to preserve equal results
among different newspaper printing presses. Furthermore we proposed the use of
appropriate measurement instruments, combined with print consistency in terms of
repeatability which are the fundamental requirements to obtain predictable
newspaper print quality.
• We have tried to answer the question: “Is there a relation between halftone
measurements from densitometers (converted into tone value with the Murray-
Davies-equation) and halftone measurements from dot meters on newspaper print?"
Our statistical analysis showed that due to large uncertainty of the estimated
parameters, the model does not accurately describe the relation between the two
measurement technologies. This can be explained by the poor repeatability
performance for dot meters applied in newspaper print. None of the three dot
meters applied in the project fulfilled the requirement of 2% tolerance deviation
(note: these are requirements which are not defined in an official standard). Dot
meters are originally developed for measuring printing plates only and are not
recommended for halftone measurements on paper substrates in newspaper
printing.
• Process control and print quality is often determined according to measurement

data obtained by different colour measurement instruments used in a production
workflow. We have demonstrated the effect of colour measurement variability
(precision and accuracy) in a colour managed printing workflow and in particular
the result of inter-model agreement measuring colours on paper substrates has been
reviewed. Moreover, in order to minimize and control the colour variations in a
colour managed printing workflow, only calibrated and certified instruments within
one certain instrument family (same model, same design of instrument from the
same manufacturer with the same parameters) should be used.
• In order to improve the colorimetric performance and inter-instrument and inter-

model agreement a method of characterizing measurement instruments using a
colorimetric regression technique directly on the measurement output values in the
CIELAB colour space has been demonstrated. The proposed method shows
significant improvements for measurements taken with different instruments in
reflective light. Furthermore, there is a potential to apply the proposed method in
110
Discussion and Conclusion Summary of Contribution
emission measurements too, using different models of spectrophotometers,

spectrocolorimeters and colorimeters.
Although we did not demonstrate ground-breaking new technology or algorithms we made

several novel contribution useful for the graphic arts and printing industry including the
demonstration of methods and procedures to ensure predictable print quality. Moreover, in
our contributions we proposed a method for the media community to control the colour
measurement uncertainties by reducing the colour differences of colour measurement
performed and simultaneously improve the colorimetric performance and inter-instrument
and inter-model agreement.
111
Summary of Contribution Discussion and Conclusion
112
Discussion and Conclusion Conclusion
4.3 Conclusion
The major goal of this work was to develop a methodology and procedures that could be
used as an application independent framework for those involved in the process of print
quality assessment. On first sight, one might assume that assessing print quality in a colour
managed printing workflow is a rather trivial task where the measurement results lead to
numerical quality estimation. However, it is more complex than it seems. A number of
factors are affecting the appearance of print. The specifications and parameters to which
the printing process has to be calibrated are often incomplete applied and the appropriate
print assessment methods are often absent.
In the present work we have demonstrated the implementation of standardization
methods and procedures in a colour managed printing workflow to ensure predictable print
quality. We found significant measurement variation among measurement instruments
used in the graphic arts and printing industry in terms of accuracy and reproducibility
performance. Hence, we proposed a solution to correct the systematic errors of the
instruments by applying a regression technique directly on the measurement output values
in the CIELAB colour space to improve the colorimetric performance and hence the inter-
instrument and inter-model agreement. The proposed method has shown significant
improvement for measurements obtained with different instruments in reflective light.
In conclusion, we strongly believe that the methodology and procedures developed
as a result of the project can be used as an application independent framework for those
involved in the process of print quality assessment. Further, with the results we have
obtained, we believe to contribute in finding solutions for the graphic arts and printing
industry to strengthen their position against other media and other global competitors in
terms of consistent and predictable print quality and cost efficient print production.
113
Conclusion Discussion and Conclusion
114
Discussion and Conclusion Perspectives
4.4 Perspectives
In the course of our research work, we have identified multiple fields for further research
in the area of colour managed printing workflow, print quality and print quality
assessment. Our thoughts about potential future research areas that should be looked into
and ideas for further improvements of the presented methodology and procedures can be
divided into three categories.
Measurement technologies: Although the proposed measurement instrument
characterization model presented in Paper G [133] showed remarkable results in terms of
reducing the measurement variations and improving the colorimetric accuracy between
different measurement instruments the method needs some further investigations. For
example the performance on different paper substrates (e.g. glossy paper, newspaper) and
material (e.g. plastic, textile, aluminium, glass) could be analyzed, the method could be
extended and tested on emission measurements using different models of measurement
devices which are used in the field. With the increasing demand for PSO certifications
which requires ambient light measurements in the prepress and press room, the
performance and reliability of the measurement instruments used needs to be investigated
and perhaps a correction model should be applied to improve the inter-instrument
agreement. An interesting aspect is to compare our model with the method applying the
correction in the spectral domain as for example proposed by Berns and Petersen [17]. It
can be speculated that a combination of the two methods could further improve the
colorimetric performance and the inter-instrument and inter-model agreement.
Colour difference and print quality metrics: The ultimate goal of implementing
standardization procedures and operating with specification conformance is the print
quality control according to numbers. Although the current ISO standards define certain
tolerance settings for process control for the production of half-tone colour separations,
proofs and production prints, the control system is based on colour patches to be measured
independently of the document content (colours in high-frequency and low-frequency
images) to be printed. In other words, the production tolerances are print content and
application independent. However, visual difference between prints from different printing
presses (or proof and print) can be perceived for certain colour shades even though the
print production on both presses are within the given tolerances. A colour control system
could be considered where the printed content automatically is related to the tolerances
according to the perceived colour difference. Suppose printing a catalogue with low-
115
Perspectives Discussion and Conclusion
frequency images (e.g. with large uniform areas) the requirements for the print variations
should be reduced compared to printed images that have a high-frequency characteristic.
Colour difference metrics for image quality assessment has been widely used for
various applications. However, there is seldom a strong correlation between the objective
evaluation and the visual assessment. Furthermore, the interpretation of the complete
image quality assessment considering the colour difference calculation depends on the
application and the acceptance. The acceptability threshold considering print quality
assessment is a vaguely defined concept and one that depends strongly on application and
industry. Moreover, a further quality metric, which could be considered more suitable for
the heat-set web offset and newspaper printing process, is a tolerance threshold.
Environmental management: As discussed in the thesis print standardization has
a number of goals such as being consistent and predictable in terms of print quality,
increasing the production efficiency and reducing costs. A further important aspect to
consider is the reduction of print waste and pollution. Pollution, global warming and
climate change have become very serious issues and even though a large number of
different research projects worldwide have been carried out aiming for environmental
management (e.g. ISO 14001 [89], Forest Stewardship Council (FSC)14) there is still a
huge need for addressing further areas contributing to the reduction of pollution and
preserving the environment and climate. The graphic arts industry has taken these issues
seriously, in particular concerning the development of new print technologies and
processes, and the increased use of recycled paper. Nevertheless, the industry’s awareness
of the importance of pollution reduction (in terms of saving energy, reducing volatile
organic compounds (VOC), and reducing waste could indeed be increased. Furthermore,
introducing colour management and standardization of the print production processes, a
further positive impact on environment and climate could be obtained.
Further research in colour management, standardization and PSO implementation
help to reduce the amount of ink and paper substrate used in print reproduction by
controlling the total ink coverage in printing on paper substrate. This should in turn help
simplify paper recycling due to reduced ink layers on paper. Research in colour
management will also help in eliminating hard copy proofing systems in the future by
introduction of softcopy proofing using high end energy efficient colour displays. The
overall effect of standardization will be to increase efficiency in terms of, increasing
14
Forest Stewardship Council is an independent, non-governmental, not-for-profit organization
established to promote the responsible management of the world’s forests, http://www.fsc.org/
116
Discussion and Conclusion Perspectives
competence among the graphic art engineers to reduce waste of paper, ink, and other
consumables.
117
Perspectives Discussion and Conclusion
118
References
References
[1] E. Aasen, Ø. Danielsen, and A. J. Bovolden, "Halftone measurements in newspaper
print," BSc, Gjøvik University College, Norway, 2002.
[2] R. M. Adams II, A. Sharma, and J. J. Suffoletto, Color Management Handbook: A
Practical Guide. Pittsburgh, PA, US: PIA/GATF Press 2008.
[3] M. Andersson and O. Norberg, "Color measurements on prints containing
fluorescent whitening agents," in Proc. SPIE: Color Imaging XII: Processing,
Hardcopy, and Applications, 2007, Vol. 6493.
[4] ASTM E284-08, Standard Terminology of Appearance, American Society for
Testing and Materials West Conshohocken, PA, USA, 2008.
[5] ASTM E308-08, Standard Practice for Computing the Colors of Objects by Using
the CIE-System, American Society for Testing and Materials, West Conshohocken,
PA, 2008.
[6] ASTM E1336-96, Standard Test Method for Obtaining Colorimetric Data From a
Visual Display Unit by Spectroradiometry, American Society for Testing and
Materials, West Conshohocken, PA, 2003.
[7] ASTM E1347-06, Standard Test Method for Color and Color-Difference
Measurement by Tristimulus Colorimetry, American Society for the Testing of
Materials, West Conshohocken, PA, 2006.
[8] ASTM E1455-03, Standard Practice for Obtaining Colorimetric Data from a
Visual Display Unit Using Tristimulus Colorimeters, American Society for the
Testing of Materials, West Conshohocken, PA, 2003.
[9] ASTM E2214-08, Standard Practice for Specifying and Verifying the Performance
of Color Measuring Instruments, American Society for the Testing of Materials,
West Conshohocken, PA, 2008.
[10] A. Bakke, I. Farup, and J. Hardeberg, "Predicting the Performance of a Spatial
Gamut Mapping Algorithm," in Color Imaging XIV: Displaying, Hardcopy,
Processing, and Applications, San Jose, CA, USA, 2009, Society of Photo-Optical
Instrumentation Engineers, SPIE Proceedings, 7241.
119
References
[11] A. M. Bakke, I. Farup, and J. Y. Hardeberg, "Evaluation of algorithms for the

determination of color gamut boundaries," Journal of Imaging Science and
Technology, vol. 54, pp. 050502-(11), 2010.
[12] E. Bando, J. Y. Hardeberg, and D. Connah, "Can gamut mapping quality be
predicted by colour image difference formulae?," in Human Vision and Electronic
Imaging X, ed. B. Rogowitz, T. Pappas, S. Daly, Proc. of SPIE-IST Electronic
Imaging, SPIE,, 2005, Volume 5666, pp. 180-191.
[13] C. Bartleson and F. Grum, Optical Radiation Measurement. Vol 5 – Visual
measurements: Academic Press, 1984.
[14] D. Battle, "The measurement of colour," in Colour physics for industry, R.
McDonald. 2nd: Society of Dyers and Colourists, 1997, 2, pp. 57-80.
[15] L. Bergman, "Using Multicoloured Halftone Screens for Offset Print Quality
Monitoring," Licentiate, Linköping University, Sweden, 2005.
[16] R. Berns, Billmeyer and Saltzman's principles of color technology: Wiley New
York, 2000.
[17] R. Berns and K. Petersen, "Empirical modeling of systematic spectrophotometric
errors," Color Research & Application, vol. 13, 4, pp. 243-256, 1988.
[18] F. Billmeyer Jr, "Comparative performance of color-measuring instruments,"
Applied Optics, vol. 8, 4, pp. 775-783, 1969.
[19] F. Billmeyer Jr and P. Alessi, "Assessment of color-measuring instruments," Color
Research & Application, vol. 6, 4, pp. 195-202, 1981.
[20] N. Bonnier, F. Schmitt, H. Brettel, and S. Berche, "Evaluation of spatial gamut
mapping algorithms," in Proc. of 14th Color Imaging Conference, IS&T,
Springfield, VA, 2006, IS&T, 14, pp. 56-61.
[21] J. Briggs, D. Forrest, and M.-K. Tse, "Reliability Issues for Color Measurement in
Quality Control Applications," in Proc. NIP 14: International Conference on
Digital Printing Technologies, 1998, IS&T, pp. 595-602.
[22] D. Brydges, F. Deppner, H. Kunzil, K. Heuberger, and R. Hersch, "Application of a
3-CCD color Camera for Colorimetric and Densitometric Measurements," in SPIE
Proceedings 3300, Color Imaging : Device Independent Color, Color Hardcopy
and Graphics Arts III, 1998, Citeseer, pp. 292-301.
[23] H. Büring, P. G. Herzog, and E. Jung, "Evaluation of Current Color Management
Tools: Image Quality Assessments," in CGIV -- Second European Conference on
120
References
Color in Graphics, Imaging and Vision, Aachen, Germany, 2004, IS&T, 2, pp.
459-462.
[24] Y. Chung, J. Xin, and K. Sin, "Improvement of inter instrumental agreement for
reflectance spectrophotometers," Coloration Technology, vol. 120, 6, pp. 284-292,
2004.
[25] CIE 15 Technical report, Colorimetry, Commission Internationale de l'éclairage,
Vienna, Austria, 2004.
[26] CIE 116, Industrial colour-difference evaluation, CIE Publication No 116-1995,
Commission Internationale de l'éclairage, Vienna, Austria, 1995.
[27] CIE 122, The relationship between digital and colorimetric data for computer
controlled CRT displays., CIE Publication No 122-1996, Commission
Internationale de l'éclairage, Vienna, Austria, 1996.
[28] CIE 130, Practical Methods for the Measurement of Reflectance and
Transmittance, CIE Publication No 130-1998, Commission Internationale de
l'éclairage, Vienna, Austria, 1998.
[29] CIE 142, Improvement to Industrial Color-difference Evaluation, CIE Publication
No 142-2001, Commission Internationale de l'éclairage, Vienna, Austria, 2001.
[30] CIE 156, Guidelines for the Evaluation of Gamut Mapping Algorithms, CIE
Publication No 156-2004, Commission Internationale de l'éclairage, Vienna,
Austria, 2004.
[31] CIE Commission Internationale de lÉclairage. Selected Colorimetric Tables 2010.
Available: http://www.cie.co.at/index.php/LEFTMENUE/index.php?i_ca_id=298
[32] F. Clarke, R. McDonald, and B. Rigg, "Modification to the JPC79 Colour
difference formula," Journal of the Society of Dyers and Colourists, vol. 100, 4, pp.
128-132, 1984.
[33] P. J. Clarke, "Measurement Good Practice Guide No.96, Surface Colour
Measurements," National Physical Laboratory, Teddington, Middlesex, United
Kingdom, TW11 0LW, 2006.
[34] S. Colthorpe and G. Imhoff, "CTP – Why densitometers do not work, White
paper," Centurfax Ltd and Grip Digital Inc. 1999.
[35] V. de Garcia, J. Inarejos, S. Otero, E. Perales, and F. Martínez-Verdú, "Assessment
of the Colorimetric Behaviour of Different Spectrophotometers," in Advances in
Printing and Media Technology: Proceedings of the 36th International Research
121
References
Conference of iarigai vol. 36, N. Enlund and M. Lovrecek. Stockholm, Sweden,

2009.
[36] DIN 16536-2, Prüfung von Drucken und Druckfarben der Drucktechnik–
Farbdichtemessung an Drucken, Deutsches Institut für Normung, 1995.
[37] F. Dugay, I. Farup, and J. Y. Hardeberg, "Perceptual evaluation of color gamut
mapping algorithms," Color Research & Application, vol. 33, 6, pp. 470-476, 2008.
[38] P. Engeldrum, Psychometric scaling: a toolkit for imaging systems development:
Imcotek Press, Winchester, Massachusetts, USA, 2000.
[39] P. Engeldrum, "Psychometric Scaling: Avoiding the pitfalls and hazards," in Proc.
IS&T’s PICS: Image Processing, Image Quality, Image Capture Systems
Conference, Montreal, Quebec, Canada, 2001, Citeseer, pp. 101-107.
[40] M. Enroth, "Environmental impact of printed and electronic teaching aids, a
screening study focusing on fossil carbon dioxide emissions," in Advances in
Printing and Media Technology: Proceedings of the 36th International Research
Conference of iarigai vol. 36, N. Enlund and M. Lovrecek. Stockholm, Sweden,
2009.
[41] European Color Initiative (ECI). ICC profiles from ECI 2010. Available:
http://www.eci.org/doku.php?id=en:downloads
[42] M. Fairchild, Color appearance models. Reading, Massachusetts, USA: Addison
Wesley Longman, Inc, 1998.
[43] M. Fairchild and F. Grum, "Thermochromism of ceramic reference tile," Applied
Optics, vol. 24, 21, pp. 3432-3, 1985.
[44] K. Falkenstern, N. Bonnier, H. Brettel, M. Pedersen, and F. Vienot, "Using Image
Quality Metrics to Evaluate an ICC Printer Profile.," in Proc. 18th Color Imaging
Conference, IS&T, San Antonio, TX, USA, 2010, IS\&T and SID, pp. 244-249.
[45] K. Falkenstern, N. Bonnier, M. Pedersen, H. Brettel, and F. Vienot, "Using Metrics
to Assess the ICC Perceptual Rendering Intent," in Image Quality and System
Performance VIII, San Francisco, CA, 2011, SPIE Proceedings, 7867.
[46] I. Farup, C. Gatta, and A. Rizzi, "A multiscale framework for spatial gamut
mapping," Image Processing, IEEE Transactions on, vol. 16, 10, pp. 2423-2435,
2007.
[47] G. Fechner, Elements of Psychophysics, Vol 1 (1860): (translated by H.E. Adler),
Holt, Rinehart and Winston, New York, N.Y, 1966.
122
References
[48] G. Field, Color and its reproduction: Fundamentals for the Digital Imaging and
Printing Industry, 3 ed.: Graphic Arts Technical Foundation, GATF, 2004.
[49] FOGRA. FOGRA28L.txt, Characterization data for standardized printing
conditions 2010. Available: http://www.fogra.de/
[50] B. Fraser, C. Murphy, and F. Bunting, Real world color management, 2 ed.:
Peachpit Press, Berkeley, CA, US, 2005.
[51] J. Gardner, "Comparison of Calibration Methods for Tristimulus Colorimeters,"
Journal of Research of the National Institute of Standards and Technology, vol.
112, 3, pp. 129-138, 2007.
[52] D. A. Garvin, "What Does 'Product Quality' Really Mean?," MIT Sloan
Management Review, vol. 26, 1, 1984.
[53] A. Gatt, S. Westland, and R. Bala, "Testing the softproofing paradigm," in Proc. of
12th Color Imaging Conference, Scottsdale, Arizona, 2004, Citeseer, pp. 187-192.
[54] A. Gatt, S. Westland, R. Bala, and Y. Ling, "Testing the Softproofing Paradigm II,"
AIC Colour 05 - 10th Congress of the International Colour Association, pp. 1243-
1246, 2005.
[55] G. Gescheider, Psychophysics: The Fundamentals, 3 ed. Mahwah, New Jersey:
Lawrence Erlbaum Associates, Inc., Publishers, 1997.
[56] E. Giorgianni and T. Madden, Digital color management: encoding solutions, 2nd
ed.: Wiley, 2008.
[57] E. M. Granger, "Comparison of Color Difference Data and Formulas," in Proc.
TAGA - Technical Association of the Graphic Arts, Proceedings of the 60th Annual
Meeting, San Francisco, CA, USA, 2008, pp. 191-201.
[58] P. Green, Color Management: Understanding and Using ICC Profiles. Chichester,
West Sussex PO19 8SQ, England: John Wiley & Sons Ltd, 2010.
[59] P. Green, "A smoothness metric for colour transforms," in Proc. SPIE. Color
Imaging XIII: Processing, Hardcopy, and Applications San Jose, CA, USA 2008,
6807, pp. 680-70I.
[60] P. Green, Understanding Digital Color, 2nd ed.: GATF Press, 1999.
[61] J. Grice and J. Allebach, "The print quality toolkit: An integrated print quality
assessment tool," The Journal of imaging science and technology, vol. 43, 2, pp.
187-199, 1999.
123
References
[62] J. Guild, "The colorimetric properties of the spectrum," Philosophical Transactions

of the Royal Society of London. Series A, Containing Papers of a Mathematical or
Physical Character, vol. 230, pp. 149-187, 1931.
[63] J. Handley, "Comparative analysis of Bradley-Terry and Thurstone-Mosteller
paired comparison models for image quality assessment," in Proceedings of the
Image Processing, Image Quality, Image Capture Systems Conference (PICS-01),
Montreal, Quebec, Canada, 2001, Citeseer, pp. 108-112.
[64] J. Y. Hardeberg, Acquisition and Reproduction of Color Images, Colorimetric and
Multispectral Approaches Dissertation.com, 2001.
[65] J. Y. Hardeberg, P. Nussbaum, S. Roch, and O. Panak, "Time matters in soft
proofing," ACTA GRAPHICA, Journal for Printing Science and Graphic
Communication, vol. 19, 171, pp. 1-10, 2007.
[66] J. Y. Hardeberg and S. Skarsbø, "Comparing color image quality of four digital
presses," in 11th International Printing and Graphic Arts Conference, Bordeaux,
France., 2002, 11th International Printing and Graphic Arts Conference.
[67] A. R. Hill, "How we see colour," in Colour Physics for Industry, R. McDonald.
Society of Dyers and Colourists, 1997, 8, pp. 426-513.
[68] J. Homann, Digital color management: principles and strategies for the
standardized print production: Springer-Verlag New York Inc, 2008.
[69] G. Hong, M. Luo, and P. Rhodes, "A study of digital camera colorimetric
characterization based on polynomial modeling," Color Research & Application,
vol. 26, 1, pp. 76-84, 2001.
[70] R. Hunt, Measuring colour, 3 ed. Kingston-upon-Thames, England: Fountain Press,
1998.
[71] R. Hunt, The reproduction of colour, 6 ed. Chichester, West Sussex PO19 8SQ,
England: John Wiley & Sons Ltd, 2004.
[72] ICC. 2004. Specification ICC.1:2004-10 (Profile version 4.2.0.0) Image technology
colour management — Architecture, profile format, and data structure. Available:
http://www.color.org/ICC1v42_2006-05.pdf, 30. November 2010.
[73] IEC 61966-2-5, Multimedia System and Equipment - Colour Measurement and
Management, Part 2-5: Colour management – Optional RGB colour space –
opRGB, International Electrotechnical Commission, 2007.
124
References
[74] F. Imai, R. Berns, and D. Tzeng, "A comparative analysis of spectral reflectance
estimated in various spaces using a trichromatic camera system," Journal of
Imaging Science and Technology, vol. 44, 4, pp. 280-287, 2000.
[75] J. Imai and M. Omodani, "Reasons why we prefer reading on paper rather than
displays: Studies for seeking paper-like readability on electronic paper," Journal of
Imaging Science and Technology, vol. 52, 5, pp. 1-5, 2008.
[76] Intergraf, "The Future of the European Printing Industry-in our own Hands," Facta
Consult, Intergraf, Brussel, Belgium, http://www.intergraf.eu/, 2010.
[77] ISO 2846-1, Graphic technology — Colour and transparency of printing ink sets
for four colour printing — Part 1: Sheet-fed and heat-set web offset lithographic
printing, International Organization for Standardization, Geneva, 2006.
[78] ISO 3664, Graphic technology and photography — Viewing conditions,
International Organization for Standardization, Geneva, 2000.
[79] ISO 3664, Graphic technology and photography — Viewing conditions,
[80] ISO 8254-1, Paper and board -- Measurement of specular gloss -- Part 1: 75
degree gloss with a converging beam, TAPPI method, International Organization
for Standardization, Geneva, 2009.
[81] ISO 9000, Quality management systems -- Fundamentals and vocabulary,
[82] ISO 12642-2, Graphic technology -- Input data for characterization of 4-colour
process printing -- Part 2: Expanded data set, International Organization for
Standardization, Geneva, 2006.
[83] ISO 12646, Graphic technology – Displays for colour proofing – Characteristics
and viewing conditions, International Organization for Standardization, Geneva,
2008.
[84] ISO 12647-1, Graphic technology – Process control for the production of half-tone
colour separations, proof and production prints – Part 1: Parameters and
measurement methods, International Organization for Standardization, Geneva,
1996.
colour separations, proof and production prints – Part 2: Offset printing processes,
125
References
colour separations, proof and production prints – Part 3: Coldset offset
lithography on newsprint, International Organization for Standardization, Geneva,
2005.
colour separations, proof and production prints – Part 7: Proofing processes
working directly from digital data, International Organization for Standardization,
Geneva, 2007.
[88] ISO 13656, Graphic technology -- Application of reflection densitometry and
colorimetry to process control or evaluation of prints and proofs, International
Organization for Standardization, Geneva, 2000.
[89] ISO 14001, Environmental management systems -- Requirements with guidance for
use, International Organization for Standardization, Geneva, 2004.
[90] ISO 15076-1, Image technology colour management -- Architecture, profile format
and data structure -- Part 1: Based on ICC.1:2004-10, International Organization
for Standardization, Geneva, 2005.
[91] ISO 15930-4, Graphic technology -- Prepress digital data exchange using PDF --
Part 4: Complete exchange of CMYK and spot colour printing data using PDF 1.4
(PDF/X-1a) International Organization for Standardization, Geneva, 2003.
Part 6: Complete exchange of printing data suitable for colour-managed workflows
using PDF 1.4 (PDF/X-3), International Organization for Standardization, Geneva,
2003.
Part 7: Complete exchange of printing data (PDF/X-4) and partial exchange of
printing data with external profile reference (PDF/X-4p) using PDF 1.6
Part 8: Partial exchange of printing data using PDF 1.6 (PDF/X-5) International
[95] ISO 23603, Standard method of assessing the spectral quality of daylight
simulators for visual appraisal and measurement of colour, International
126
References
[96] ISO/IEC 13660, Information technology -- Office equipment -- Measurement of

image quality attributes for hardcopy output -- Binary monochrome text and
graphic images, International Organization for Standardization, Geneva, 2001.
[97] ISO/TS 10128, Graphic technology -- Methods of adjustment of the colour
reproduction of a printing system to match a set of characterization data,
[98] K. Johansson, P. Lundberg, and R. Ryberg, A guide to graphic print production,
2nd ed. Hoboken, New Jersey: Wiley & Sons, Inc., 2007.
[99] T. Johnson, Colour management in graphic arts and publishing. Surrey, UK: Pira
International, 1996.
[100] T. Johnson, "Methods for characterizing colour scanners and digital cameras,"
Displays, vol. 16, 4, pp. 183-191, 1996.
[101] H. Kang, Color technology for electronic imaging devices: SPIE-International
Society for Optical Engineering, 1997.
[102] N. Katoh, "Practical method for appearance match between soft copy and hard
copy," in Device-Independent Color Imaging, Proceeding of SPIE, 1994, SPIE,
2170, pp. 170-181.
[103] B. W. Keelan, Handbook of Image Quality: Characterization and Prediction. New
York NY 10016: Marcel Dekker, Inc., 2002.
[104] Y. Kipman, "Image quality metrics for printers and media," in IS&T’s Image
Processing, Image Quality, Image Capture, Systems Conference, Portland, Oregon,
USA, 1998, The Society for Imaging Science and Technology, 1, pp. 183-187.
[105] H. Kipphan, Handbook of print media: technologies and production methods:
Springer-Verlag Berlin Heidelberg, 2001.
[106] R. Kuehni, "CIEDE2000, milestone or final answer?," Color Research &
Application, vol. 27, 2, pp. 126-127, 2002.
[107] R. Kuehni, "Towards an improved uniform color space," Color Research &
[108] H. Kuenzli, F. Deppner, K. Heuberger, and Y. Jiang, "Mini-Targets: A New
Dimension In Print Quality Control," in Proc. Color imaging: device-independent
color, color hardcopy, and graphic arts III, San Jose, California, 1998, Society of
Photo Optical, 3300, pp. 286-291.
127
References
[109] S. Leckner and S. Nordqvist, "Soft Proofing using LCDs-Case Newspaper

Workflow," in Proc. TAGA - Technical Association of the Graphic Arts,
Proceedings of the 54th Annual Meeting, 2002, TAGA; 1998, pp. 367-378.
[110] S. Lindberg, D. C. Donderi, and N. Pauler, "Image quality of four digital printing
methods compared to offset and flexographic printing. In Advances in Color
Reproduction," in Proceedings of the 28th International Iarigai Research
Conference, chapter 4.1. , Sewickley, PA, US, 2001, GATF Press.
[111] M. Luo, G. Cui, and B. Rigg, "The development of the CIE 2000 colour difference
formula: CIEDE2000," Color Research & Application, vol. 26, 5, pp. 340-350,
2001.
[112] M. Luo, G. Cui, and B. Rigg, "Further comments on CIEDE2000," Color Research
& Application, vol. 27, 2, pp. 127-128, 2002.
[113] D. MacAdam, "Visual sensitivities to color differences in daylight./. opt," Soc.
Amer, vol. 32, pp. 247-274, 1942.
[114] M. Mahy, L. Van Eycken, and A. Oosterlinck, "Evaluation of uniform color spaces
developed after the adoption of CIELAB and CIELUV," Color Research and
[115] G. Marcu, "Color Quality in Desktop Printing," in IS&T’s PICS: Image
Processing, Image Quality, Image Capture Systems Conference, Portland, Oregon,
USA, 2000, IS&T, Tutorial Notes.
[116] G. Marcu, "Factors Governing Print Quality in Color Prints," in Proc. IS&T’s
PICS: Image Processing, Image Quality, Image Capture Systems Conference,
Springfield, VA, USA, 1998, IS&T, pp. 221-225.
[117] D. Martin, J. O'Neill, T. Colombino, F. Roulland, and J. Willamowski, "'Colour,
itís Just a Constant Problem': An Examination of Practice, Infrastructure and
Workflow in Colour Printing," in From CSW to Web 2.0: European Developments
in Collaborative Design, Computer Supported Cooperative Work, D. Randall and
P. Salembier. London: Springer-Verlag, 2010, 2, pp. 21-42.
[118] Microsoft Corporation. Windows Color Quality Specifications for Printer OEMs.
Part of Microsoft Hardware Quality Labs (WHQL)’s Windows Color Quality Test
Kit. 2001. Available: http://www.microsoft.com/hwdev/tech/color/ColorTest.asp.
[119] Å. Moberg, C. Borggren, G. Finnveden, and S. Tyskeng, "Effects of a total change
from paper invoicing to electronic invoicing in Sweden," Report from the KTH
Centre for Sustainable Communications TRITA-SUS, vol. 3, 2008.
128
References
[120] Å. Moberg, M. Johansson, G. Finnveden, and A. Jonsson, Screening environmental

life cycle assessment of printed, web based and tablet e-paper newspaper: Centre
for Sustainable Communications, Royal institute of technology (KTH), 2007.
[121] J. Morovic, Color Gamut Mapping. Chichester: John Wiley & Sons, Ltd, 2008.
[122] J. Morovic, "Colour gamut mapping," in Colour Engineering Achiving Device
Independent Colours, P. Green and L. MacDonald. Chichester, West Sussex PO19
8SQ, England: John Wiley & Sons Ltd, 2002, pp. 297-317.
[123] J. Morovic, "Gamut mapping," in Digital Color Imaging Handbook. , G. Sharma.
CRC Press, 2003, Chapter 10, pp. 639-685.
[124] J. Morovic and M. Luo, "The fundamentals of gamut mapping: A survey," Journal
of Imaging Science and Technology, vol. 45, 3, pp. 283-290, 2001.
[125] J. Morovic and M. Luo, "The Pleasantness and Accuracy of Gamut Mapping
Algorithms," in ICPS Conference 1998, 2, pp. 39-43.
[126] J. Morovic and P. Nussbaum, "Factors affecting the appearance of print on opaque
and transparent substrates," Journal of Imaging Science and Technology, vol. 47, 6,
pp. 554-564, 2003.
[127] A. Murray, "Monochrome reproduction in photoengraving," Journal of the
Franklin Institute, vol. 221, 6, pp. 721-744, 1936.
[128] NADA. 2010. Support – Lineær avisproduksjon – kort og godt. [Guideline].
Available: http://www.nada.no/support/linprod/index.php, retrieved 13. April 2010.
[129] O. Norberg, P. Westin, S. Lindberg, M. Klaman, and L. Eidenvall, "Comparison of
print quality between digital and traditional technologies," in In Proceedings of
DPP2001: International Conference on Digital Production Printing and Industrial
Applications, 2002, pp. 380-385.
[130] P. Nussbaum and J. Y. Hardeberg, "Print Quality Evaluation and Applied Colour
Management in Coldset Offset Newspaper Print," in Proc. TAGA - Technical
Association of the Graphic Arts, Proceedings of the 60th Annual Meeting, San
Francisco, CA, USA, 2008, pp. 372-391.
[131] P. Nussbaum and J. Y. Hardeberg, "Print Quality Evaluation and Applied Colour
Management in Coldset Offset Newspaper Print," Color Research & Application,
Article first published online:March 8th 2011, DOI: 10.1002/col.20674
[132] P. Nussbaum and J. Y. Hardeberg, "Print quality evaluation and applied colour
management in heat-set web offset," in Advances in Printing and Media
Technology: Proceedings of the 33rd International Research Conference of
129
References
iarigai. vol. 33, N. Enlund and M. Lovrecek. Acta Graphica Publishers, 2006, pp.
399-411.
[133] P. Nussbaum, J. Y. Hardeberg, and F. Albregtsen, "Regression based
characterization of Color Measurement instruments in printing applications," in
Electronic Imaging, Color Imaging XVI: Displaying, Processing, Hardcopy, and
Applications, San Francisco, CA, 2011, SPIE Proceedings, 7866.
[134] P. Nussbaum, J. Y. Hardeberg, and S. E. Skarsbø, "Print quality evaluation for
governmental purchase decisions," in Advances in Printing Science and
Technology: Proceedings of the 31st International Research Conference of iarigai
vol. 31, M. Lovrecek. Acta Graphica Publishers, 2004, pp. 189-200.
[135] P. Nussbaum and J. Morovic, "Factors Affecting The Appearance Of Print On
Opaque Substrates," in Proc. TAGA - Technical Association of the Graphic Arts,
Proceedings of the 55th Annual Meeting, Montreal, Canada, 2003, pp. 47-62.
[136] P. Nussbaum, A. Sole, and J. Y. Hardeberg, "Analysis of colour measurement
uncertainty in a colour managed printing workflow," Journal of Print and Media
Technology Research, iarigai, Accepted for publication.
[137] P. Nussbaum, A. Sole, and J. Y. Hardeberg, "Consequences of using a number of
different color measurement instruments in a color managed printing workflow," in
Proc. TAGA - Technical Association of the Graphic Arts, Proceedings of the 61st
Annual Meeting, New Orleans, LA, USA, 2009, pp. 324-346.
[138] Y. Ohno, "CIE Fundamentals for color measurements," in Proc. NIP 16:
International Conference on Digital Printing Technologies, Vancouver, CA, 2000,
IS&T, pp. 540-545.
[139] N. Ohta and A. Robertson, Colorimetry: Fundamentals and Applications: J. Wiley,
2005.
[140] Z. Pan, Y. Noyes, J. Hardeberg, L. Lee, and G. Healey, "Color scanner
characterization with scan targets of different media types and printing
mechanisms," in Color Imaging: Device-Independent Color, Color Hardcopy, and
Graphic Arts VI, San Jose, CA, 2001, SPIE Proceedings, 4300, pp. 58-63.
[141] O. Panak, P. Nussbaum, and J. Y. Hardeberg, "Colour Memory Match Under
Disparate Viewing Conditions," in Fifteenth Color Imaging Conference: Color
Science and Engineering Systems, Technologies, and Applications, IS&T,
Albuquerque, New Mexico, 2007, pp. 325-330.
130
References
[142] D. R. Pant and I. Farup, "Riemannian Formulation of the CIEDE2000 Color

Difference Formula," in Proc. 18th Color Imaging Conference, IS&T, San
Antonio, TX, USA, 2010, IS&T, pp. 103-108.
[143] M. Pedersen, Internationalt Standardiseret Grafisk Produktion ISO 12647: Grafisk
Arbejdsgiverforening, 2008.
[144] M. Pedersen and S. A. Amirshahi, "Framework for the Evaluation of Color Prints
Using Image Quality Metrics," in Proc. 5th European Conference on Colour in
Graphics, Imaging, and Vision (CGIV), Joensuu, Finland, 2010, IS&T, pp. 75-82.
[145] M. Pedersen, N. Bonnier, J. Y. Hardeberg, and F. Albregtsen, "Attributes of a New
Image Quality Model for Color Prints," in Proc. 17th Color Imaging Conference,
IS&T, Albuquerque, NM, USA, 2009, IS&T, pp. 204-209.
[146] M. Pedersen, N. Bonnier, J. Y. Hardeberg, and F. Albregtsen, "Attributes of image
quality for color prints," Journal of Electronic Imaging, vol. 19, 1, pp. 011-016,
2010.
[147] M. Pedersen, N. Bonnier, J. Y. Hardeberg, and F. Albregtsen, "Estimating Print
Quality Attributes by Image Quality Metrics," in Proc. 18th Color Imaging
Conference, IS&T, San Antonio, TX, USA, 2010, IS&T, pp. 68-73.
[148] M. Pedersen, N. Bonnier, J. Y. Hardeberg, and F. Albregtsen, "Validation of
Quality Attributes for Evaluation of Color Prints," in Proc. 18th Color Imaging
Conference, IS&T, San Antonio, TX, USA, 2010, IS&T, pp. 74-79.
[149] G. Radencic, E. Neumann, and M. Bohan, "Spectrophotometer inter-instrument
agreement on the color measured from reference and printed samples," in Proc.
TAGA - Technical Association of the Graphic Arts, Proceedings of the 60th Annual
Meeting, San Francisco, CA, 2008, pp. 232-251.
[150] R. Rasmussen, "Measuring Color Printer Image Quality," in Tutorial Notes,
IS&T/SID 9th Color Imaging Conference, Springfield, VA, USA, 2000, course
SC4.
[151] D. C. Rich, "Instruments and methods for colour measurement," in Colour
engineering: achieving device independent colour, P. J. Green and L. W.
MacDonald. Wiley, 2002.
[152] D. C. Rich, Y. Okumura, and V. Lovell, "The Effect of Spectrocolorimeter
Reproducibility on a Fully Color-Managed Print Production Workflow," in 4th
European Conference on Color in Graphics, Imaging, and Vision (CGIV),
Barcelona, Spain 2008, Society for Imaging Science and Technology, pp. 77-79.
131
References
[153] T. Rijgersberg. Calculation from the original experimental data of the CIE 1931
RGB standard observer spectral chromaticity co-ordinates. 2009. Available:
http://www.cis.rit.edu/mcsl/research/1931.php
[154] D. Roberts, Signals and perception: the fundamentals of human sensation:
Palgrave, 2002.
[155] A. Robertson, "Diagnostic performance evaluation of spectrophotometers," in
Advances in standards and methodology in spectrophotometry. vol. 2, B. C. a. M.
K.D. Elsevier Science Publisher B.V., Amsterdam, 1987, pp. 277-286.
[156] S. Roch, J. Y. Hardeberg, and P. Nussbaum, "Effect of time spacing on the
perceived color," in Proc. SPIE: Color Imaging XII: Processing, Hardcopy, and
Applications 2007, The International Society for Optical Engineering, Color
Science, 6493, pp. 185-195.
[157] J. Rodgers, K. Wolf, N. Willis, D. Hamilton, R. Ledbetter, and C. Stewart, "A
comparative study of color measurement lntstrumentation," Color Research &
[158] D. Romano, "The Image Analyzer – A True Dot Area Meter?," in Proc. TAGA -
Technical Association of the Graphic Arts, Proceedings of the 51st Annual
Meeting, Vancouver, Canada, 1999, pp. 318-334.
[159] J. Schanda, Colorimetry: Understanding the CIE system: Wiley-Interscience, 2007.
[160] J. Schanda, G. Eppeldauer, and G. Sauter, "Tristimulus Color Measurement of self-
Luminous Sources," in Colorimetry: Understanding the CIE system, J. Schanda.
Wiley-Interscience, 2007, pp. 135-157.
[161] K. Schläpfer, Farbmetrik in der Reproduktionstechnik und im Mehrfarbendruck, 2
ed.: UGRA, 1993.
[162] U. Schmitt and F. Dolezalek, Ugra/FOGRA Media Wedge CMYK V 2.0, Munich,
2004.
[163] M. Scorer, M. Luo, and B. Rigg, "Performance monitoring of colour measuring
instruments," Coloration Technology, vol. 118, 4, pp. 149-153, 2002.
[164] J. Seymour, "Color measurement with an RGB camera," in Proc. TAGA -
Technical Association of the Graphic Arts, Proceedings of the 61st Annual
Meeting, New Orleans, LA, USA, 2009, pp. 79-105.
[165] J. Seymour, "How Many Δ Es Are There in a ΔD?," in Proc. TAGA - Technical
Association of the Graphic Arts, Proceedings of the 59th Annual Meeting,
Pittsburgh, PA, USA, 2007, pp. 311-348.
132
References
[166] A. Sharma, "Methodology for Evaluating the Quality of ICC Profiles-Scanner,

Monitor, and Printer," Journal of Imaging Science and Technology, vol. 50, p. 469,
2006.
[167] A. Sharma, Understanding Color Management. Thompson Delmar Learning: New
York, 2004.
[168] A. Sharma and P. Fleming, "Measuring the Quality of ICC Profiles and Color-
Management Software," Seybold Report, vol. 2, 19, p. 3, 2003.
[169] G. Sharma, Digital color imaging handbook: New York, CRC, Press, 2003.
[170] G. Sharma and H. Trussell, "Digital color imaging," IEEE Transactions on Image
Processing, vol. 6, 7, pp. 901-932, 1997.
[171] D. Silverstein and J. Farrell, "The relationship between image fidelity and image
quality," in Proc. IEEE Int. Conference Image Processing, Lausanne, Switzerland,
1996, IEEE, 1, pp. 881-884.
[172] A. Sole, P. Nussbaum, and J. Y. Hardeberg, "Implementing ISO 12646 standards
for soft proofing in a standardized printing workflow according to PSO," in
Advances in Printing and Media Technology: Proceedings of the 37th International
Research Conference of iarigai, vol. 37, N. Enlund and M. Lovrecek. International
Association of Research Organizations for the Information, Media and Graphic
Arts Industries, 2010, pp. 215-226.
[173] M. Solli, M. Andersson, R. Lenz, and B. Kruse, "Color measurements with a
consumer digital camera using spectral estimation techniques," Image Analysis, pp.
105-114, 2005.
[174] T. Steder, M. R. Luo, and Li;Changjun, "Training Data Selection Study for Surface
Colour Measurement Data Correlation," in CGIV 2008 -- Fourth European
Conference on Colour in Graphics, Imaging, and MCS/08 Vision 10th
International Symposium on Multispectral Colour Science, Terrassa – Barcelona,
España, 2008, Society for Imaging Science and Technology.
[175] S. Stevens, "On the theory of scales of measurement," Science, vol. 103, 2684, pp.
677-680, 1946.
[176] S. Stevens, Psychophysics: Wiley New York, 1975.
[177] M. Stokes, "The impact of color management terminology on image quality," in
PICS Conference, Portland, Oregon, USA, 1998, Society for Imaging Science &
Technology, pp. 216-220.
133
References
[178] M. Stokes, "Industry adoption of color management systems," in Proceedings of

the 8th Concress of the International Colour Association, AIC Color 97, Kyoto,
Japan, 1997, Vol I, pp. 126-131.
[179] M. Stokes, M. Fairchild, and R. Berns, "Colorimetrically quantified visual
tolerances for pictorial images," in Proc. TAGA - Technical Association of the
Graphic Arts, Proceedings of the 44th Annual Meeting, Williamsburg, VA, USA,
1992, pp. 757-777.
[180] B. Streckel, B. Steuernagel, E. Falkenhagen, and E. Jung, "Objective print quality
measurements using a scanner and a digital camera," in DPP2003: IS&Ts
International Conference on Digital Production Printing and Industrial
Applications, Barcelona, Spain, 2003, 5, pp. 145-147.
[181] L. Thurstone, "A law of comparative judgment," Psychological review, vol. 34, 4,
pp. 273-286, 1927.
[182] Tobias Associates Inc. 2007. Reflection Densitometry. Available:
http://ourworld.compuserve.com/homepages/tobiasinc/TAPages/bulletin3.htm,03.0
3.2007.
[183] W. Torgerson, "A law of categorical judgment," Consumer behavior, pp. 92-93,
1954.
[184] UGRA. (2009, 6. July 2010). PSO certification. Available:
http://www.ugra.ch/pso-certification.phtml
[185] A. Valberg, Light vision color: Wiley, 2005.
[186] J. F. Verrill, P. J. Clarke, and J. O'Halloran, "Study of Improved Methods for
Absolute Colorimetry, Fifth Report, Study of Colorimetric Errors in Industrial
Instruments," Centre for Optical and Environmental Metrology, National Physical
Laboratory, Teddington, Middlesex, United Kingdom, TW11 0LW, 1997.
[187] H. Völz, Industrial color testing: fundamentals and techniques, 2nd ed.: Wiley,
2001.
[188] B. Wandell, Foundations of vision: Sinauer Associates, 1995.
[189] Z. Wang and A. Bovik, Modern image quality assessment: Morgan & Claypool
Publishers, 2006.
[190] Z. Wang, A. Bovik, H. Sheikh, and E. Simoncelli, "Image quality assessment:
From error visibility to structural similarity," Image Processing, IEEE Transactions
on, vol. 13, 4, pp. 600-612, 2004.
134
References
[191] S. Westland and C. Ripamonti, Computational colour science using MATLAB:

Wiley, 2004.
[192] W. D. Wright, "A re-determination of the trichromatic coefficients of the spectral
colours," Transactions of the Optical Society, vol. 30, p. 141, 1929.
[193] M. Wroldsen, P. Nussbaum, and J. Y. Hardeberg, "Densitometric and Planimetric
Measurement Techniques for Newspaper Printing," in Proc. TAGA - Technical
Association of the Graphic Arts, Proceedings of the 59th Annual Meeting,
Pittsburgh, PA, USA, 2007, pp. 273-290.
[194] M. Wroldsen, P. Nussbaum, and J. Y. Hardeberg, "Densitometric and Planimetric
Measurement Techniques for Newspaper Printing " Journal TAGA - Technical
Association of the Graphic Arts, vol. 4, pp. 101-115, 2008.
[195] D. R. Wyble. (2010, 24. August 2010). Color Measurement - Introduction,
Historical Perspective, Definitions and Terminology, Components of a
Spectrophotometer, Light Source, Detector, Dispersing Element. Available:
http://encyclopedia.jrank.org/articles/pages/1246/Color-Measurement.html
[196] D. R. Wyble and D. Rich, "Evaluation of methods for verifying the performance of
color-measuring instruments. Part I: Repeatability," Color Research & Application,
vol. 32, 3, pp. 166-175, 2007.
[197] D. R. Wyble and D. Rich, "Evaluation of methods for verifying the performance of
color-measuring instruments. Part II: Inter-instrument reproducibility," Color
Research & Application, vol. 32, 3, pp. 176-194, 2007.
[198] G. Wyszecki and W. Stiles, Color science, Concepts and Methods, Quantitative
Data and Formulae, 2 ed.: Wiley New York, 2000.
[199] S. Yendrikhovskij, "Image quality: Between science and fiction," in In Proc.
IS&T’s PICS Conference, Savannah, Georgia, US., 1999, IS&T, pp. 173-178.
[200] J. Zwinkels, "Colour-measuring instruments and their calibration," Displays, vol.
16, 4, pp. 163-171, 1996.
135
References
136
PART II
INCLUDED PAPERS
137
138
Paper A
Paper A
Peter Nussbaum, Jon Yngve Hardeberg and Sven Erik Skarsbø
Published
In Advances in Printing Science and Technology:
Proceedings of 31st International iarigai Research Conference,
Volume 31, pp. 189-200,
M. Lovreček, Ed.,
Acta Graphica Publishers,
2004.
139
Paper A
140
Paper A
Peter Nussbaum, Jon Y. Hardeberg and Sven Erik Skarsbø

P.O.Box 191
N-2802 Gjøvik
Norway
peter.nussbaum@hig.no
jon.hardeberg@hig.no
sven.skarsbo@hig.no
Abstract
Potential costumers within the digital printer market have various demands considering
their requirements with regards to desired print quality. This paper aims to investigate the
print quality according to predefined quality factors to determine the appropriate printing
equipment. In particular the study describes the methods and results of a research project
conducted by researchers at the Norwegian Color Research Laboratory for the Norwegian
Government Administration Service. The objective of the project was to develop methods
and procedures to perform test prints from digital printers including the evaluation and
assessment of the print quality according to predefined quality factors. The results of the
study indicate the performance of various digital printers in terms of their obtained print
quality, which will be useful for the governmental purchase decisions.
1 Introduction
In 1997 the design program for the Norwegian Ministries was established, and a design
handbook was produced (Kulturdepartementet, 2000). The Ministry for Culture and
Church Affairs is responsible for the design program and the ministry has delegated the
responsibility for the administration, the daily running and the development of the design
program to the Norwegian Government Administration Service (NGAS). NGAS is also
technical editor for governmental publications, and supports all of the ministries in the
printing of publications. Amongst other things, the design handbook describes and defines
the colours used in the ministries logos (National Coat-of-Arms). In order to maintain
141
Paper A
control over colour reproduction of these logos it was decided to use pre-printed
letterheads; letterheads produced by offset printing.
Originally, the three colours of the national logo have been specified with regard to
PMS (Pantone Matching System). Although this printing technique ensures the colour
accuracy of the colour logo, it is an expensive and time consuming process. On the other
hand, after several years of market hesitation, digital presses have now become common
and in today’s printing market, where flexibility, variable content, shorter lead times, and
on demand publishing, are being demanded, digital presses represent an attractive
supplement to conventional offset presses (Hardeberg and Skarsbø, 2002). Consequently,
the Ministry for Transport and Communications has decided to embrace this new
technology and replace the conventional offset process by digital printing systems. In order
to maintain the colour reproduction demands, as outlined in the design handbook, the
Norwegian Color Research Laboratory at Gjøvik University College in collaboration with
NGAS has carried out a project to develop methods to evaluate digital printers.
Consequently, the aim of the presented work has been to develop methods and quality
factors to evaluate digital printers in terms of their potential impact on print quality, in
accordance with the quality requirements of the NGAS. Furthermore, the development of
digital test documents and test procedures to obtain the evaluation results is an important
part of this study. Although other aspects, such as the commercial and the technical point
of view, might be considered for the acquisition of the appropriate equipment, the results
of the printer evaluation will be used by the NGAS to determine the final purchase
decisions. We also believe that not only the results but also the methodology will be of
interest to other target groups and who are considering the replacement of conventional
printing by digital printing.
NGAS, as the costumer of this project, nationally announced a printer quality contest
and printer manufactures were invited to participate in the test. However, due to the current
purchase process by the NGAS, the names of the participating printer manufactures and
their printing systems cannot be published in this work.
In the reminder of this study the next section gives a short overview in the field of colour
image quality. Then section three introduces the experimental method used for evaluating
the print quality in terms of the predefined quality factors. Then in section four the data
analysis performed on the experimental data is described, followed by a presentation of the
142
Paper A
results. Finally in the section five the implications of the results will be discussed and ideas
will be suggested for future work.
2 Colour Image Quality
Various studies and research have been done in the field of print quality and repeatedly it
has been concluded as a very complex issue. The subject has been discussed at various
conferences (see e.g. Stokes, 1998; Hardeberg and Skarsbø, 2002). In another study by
Marcu (2000) addresses various quality factors that determine the print quality such as
printing technology, colorant/media interaction, geometric resolution, halftoning,
separation, black generation, UCR, GCR and tone reproduction.
Principally there are two different methods to assess image quality. The first method is
by measurement, using instruments to determine values for the various quality factors.
Considering the evaluation of relative colorimetric colour reproduction of printing devices
Microsoft has proposed a quality assurance system (Microsoft Corporation, 2001).
The second method is based on observation, using psychophysical experiments to gather
the judgement of human observers. For instance, the pair comparison method is a robust
approach where observers are asked to compare the perceptual magnitude between two
stimuli (or pairs of stimuli). This method is based on Thurstone’s «law of comparative
judgement» (Handley, 2001; Gescheider, 1985).
To produce an accurate assessment of image quality the values obtained by
measurement and observation metrics can be calculated. The term metric is generally
applied to any physical and psychophysical measure of image quality. Image quality
metrics can be defined as numbers, derived from physical measurements that can be
related to the perceptions of image quality (Jacobson, 1995).
Compared to a number of other studies, whose aim is the comparison of colour image
quality of digital presses concentrating in the field of rendering complex images, the main
focus in this work is printer evaluation according to reproduction requirements of the
Norwegian ministries logo (National Coat-of-Arms). Hence, considering the printer
evaluation in this work, NGAS has most emphasis the colour logo reproduction (Figure 1).
143
Paper A
Figure 1: The design handbook describes and defines the colours used in the Norwegian
ministries logo (National Coat-of-Arms).
3 Experimental Method
Considering the methods to assess the digital print quality in this study we have proposed a
number of quality factors. According to the presented list of appropriate quality factors
NGAS has determined the most significant quality factors including their weights. Then
the selected quality factors have been divided into three assessment categories, namely,
“visual logo assessments” (Section 3.1), “general print quality assessments” (Section 3.2)
and finally the “copy quality assessments” (Section 3.3).
To complete the final ranking the results of the three categories have been weighted
once again. According to the priorities of the quality requirements defined by NGAS the
category “visual logo assessments” has been weighted by 50%, the category “general print
quality assessments” by 30% and finally the third category “copy quality assessments” by
20%.
The quality evaluation and the final ranking were done at the Color Research
Laboratory at Gjøvik University College, using psychophysical experiments (expert panel)
and quantitative analyses based on measurements. The psychophysical evaluations were
performed to determine the colour logo reproduction in terms of colour match, visual
resolution, surface texture, logo alignment and artefacts. The quantitative analyses
included text quality, colour gamut, repeatability and register. The next three sections are
describing the assessment categories including the defined quality factors.
3.1 Visual logo assessment
As mentioned in the previous section, the Color Research Laboratory has proposed a
number of quality factors to determine the colour reproduction performance of digital
printers. In the first category, which is considering the visual assessment of the colour logo
reproduction, Table 1 lists the selected quality factors and the corresponding weight
144
Paper A
defined by the NGAS. It can be seen that the quality factor “colour match” has absolute
highest priority.
Table 1: Quality factors within the category visual logo assessments and the
corresponding weights.
Quality factors Weight %
Colour match 50
Visual resolution 20
Surface texture 10
Logo alignment 10
Artefacts 10
The next paragraph explains the quality factors in more detail.

• Colour match: The aim of this factor is to obtain, as close as possible, a colour
match between the logo colours defined by the three Pantone colours and the test
prints. The competitors have been asked to adjust the source file with the three logo
colours according to the supplied Pantone colour patches to achieve a closest
possible colour match between the test print and the Pantone target.
• Visual resolution: Detail reproduction in the colour logo such as the «tongue in the
lions head» (Figure 1), sharp lines and smooth edges have an impact in the quality
of the logo reproduction.
• Surface texture: Observation of graininess, texture of the colours, smoothness of
the image and banding.
• Logo alignment: Positioning of the colour logo according to the requirements. The
colour logo placed on the A4 document has to be aligned centre and 10mm from
the top.
• Artefacts: Observation of unwanted effects on the print such as stripes, scratches
and marks.
A psychophysical experiment was carried out to evaluate the visual quality of the colour
logo reproduction that the printers can obtain. The test prints have been visually assessed
according to standardised illumination and viewing conditions (ISO 3664 P1). Before
performing the visual assessment the expert panel, including three members from the
NGAS and one member from the Color Research Laboratory, have conducted the
ISHIHARA Test for Colour Deficiency to confirm their appropriate colour vision.
145
Paper A
The magnitude estimation method was employed to judge each quality factor. The
four observers gave individual ratings on a scale from excellent (6) to very poor (0). The
individual marks obtained by the four experts have been collected and the average result
listed by ranking.
3.2 General print quality assessments

Within this category, four quality factors have been proposed to assess the print quality
independently of the Norwegian ministries logo. Table 2 presents the selected quality
factors and the corresponding weights. Note that NGAS has weighted all factors equal.
Table 2: Quality factors within the category general print quality assessments and the

Text quality 25
Colour gamut 25
Repeatability 25
Register 25
• Text quality: This quality factor analyses the performance of the text reproduction
including negative text in various text sizes (3 pica point up). In addition the
rendering of text on coloured background, bitmap and bitmap illustrations are further
assessed. The assessment method used is based on observation from the expert panel
rating on a scale from excellent (6) to very poor (0).
• Colour gamut: The colour gamut defines the range of colours achievable on a certain
substrate under the predefined viewing conditions. Colour difference in terms of
ΔE*ab between the logo colours (Pantone 280M, Pantone 185M and Pantone 116M)
and the corresponding logo colours printed by the competitors is part of the
assessment. A printer with a larger colour gamut than another is typically able to
reproduce more saturated colours. However considering the predefined Pantone
colours in the colour logo, it is not expected that a larger colour gamut necessarily
include the target colours because the match will be determined due to the location
of the Pantone colour and the device colour gamut within the entire colour space.
The data used in this task are based on colorimetric measurements, which have been
analysed and visualised by the icc3d application (Farup et al., 2002).
146
Paper A
• Repeatability: To determine the variation in terms of colour stability over time the
quality factor repeatability is required (Morovic and Nussbaum, 2003). Essential, a
Microsoft Word document (102 pages), representing an ordinary file including text,
images, graphs, illustrations and two colour test charts (inserted on page 6 and page
99) was the source. To quantify the short-term repeatability of the printer, the Word
document was printed five times in sequence. This results in ten measurements of
the colour test chart with an interval of 93 pages and 9 pages respectively. The
variations between the measurements are given in ΔE*ab units.
• Register: The register between individual colour image layers influences the print
quality of the colour reproduction (Field, 1999). Variations in register can result in
loss of resolution and sharpness. Figure 2 illustrates a sample of nonius register
marks (Ifra, 2002) to determine the register deviations in the print. The evaluation to
determine the register performance was carried out by the use of a microscope
according to ISO 12647-2.
Figure 2: Variations in register can be determined by the aid of nonius register marks
(Ifra, 2002).
As mentioned in the previous section NGAS is responsible for the design program (design
handbook), which describes and defines the colours used in the ministry logo (Figure 1).
The three colours of the national logo, accepted by NGAS as «the optimised logo colours»,
refer to Pantone 280M, Pantone 185M and Pantone 116M. Table 3 shows the CIELAB-
values of the three logo colours obtained by colour measurement at the Norwegian Color
Research Laboratory using a spectrophotometer, SpectroLino (2°, D50). These obtained
CIELAB-values for “optimised logo colours” are registered in the design program.
147
Paper A
Table 3: CIELAB-values (2°, D50) for “optimised logo colours” registered in the design
program.
L* a* b*
Pantone 280M 20.70 11.50 -51.10
Pantone 185M 52.70 71.90 34.00
Pantone 116M 84.70 6.90 83.20
Although this category combines visual observations and the method using instruments to
determine values for some of the quality factors, a scaling from excellent (6) to very poor
(0) has been employed.
3.3 Copy quality assessments
Finally in the last category, three quality factors have been analysed determining the
quality performance of the copy function of the printers. The evaluation method used is
based on visual assessment with individual marks given by the expert panel rating on a
scale from excellent (6) to very poor (0). The quality factors in this test and the
corresponding weights are listed in Table 4.
Table 4: Quality factors within the category copy quality assessments and the

Colour match between original print and copy 40
Copy accuracy 40
Artefacts 20
• Colour match between original print and copy: The original print for this task is the
Word document with 102 pages printed from the supplied digital file. Subsequently
a copy has been produced, by using the copy function of the printer.
• Copy accuracy: The rendering ability in the copy reproduction of text and
illustrations has been assessed.
• Artefacts: Undesirable print artefacts, which appear on the copy print such as dust,
stripes, scratches and marks have been evaluated.
148
Paper A
3.4 Test targets
As mentioned previously, NGAS, as the costumer of this project, nationally announced a

printer quality contest and printer manufactures were invited to participate in the test.
Consequently methods had to be designed, including the development of digital test
documents and test procedures, for the competitors that have applied to conduct the test.
The requirements for conducting the test were defined in four tasks. Included on a CD-
ROM each competitor has been asked to print four different test documents.
The first test document, defined in pdf-format, sRGB colour space and in size A3,
contains test images, colour charts and elements for the register control. The second target,
shown in Figure 3, was specified in pdf-format, CMYK for offset (coated paper) and in
size A3. It consists of test images, colour charts and other control elements according to
the CMYK colour space. The third document is a Microsoft Word file, A4 in size,
containing 102 pages with text, illustrations and colour test charts. This test file should
simulate a customised text document completed by the NGAS. The fourth document is an
eps-file consisting of the three Pantone colours, which represent the national colour logo
(National Coat-of-arms). The participants have been asked to adjust the Pantone colours in
the file to obtain a closest possible match between the print and the «original Pantone
patches». NGAS has provided the three Pantone colour patches Pantone 280M, Pantone
185M and Pantone 116M. Finally, the last task in the test, the competitors have been asked
to copy the test prints from the Microsoft Word file on the same equipment as the test
prints have been made. Note that the presented test documents contains test elements such
as complex images, which are not taken into consideration within the test. However, these
elements could be used in further image quality assessments.
Figure 3: The CMYK test target designed for the printer test.
149
Paper A
The competitors were allowed to choose their customized printing parameters such as
driver settings and RIP software, with the goal of achieving their best colour image quality.
Considering the substrate, NGAS has supplied all the participants with the appropriate
paper, which has been required conducting the print test. The scanning function of the
printers has not been part of the evaluation.
4 Results
In this section the results of the printer evaluation will be presented. To recap the aim of
this work is to develop methods including quality factors to the evaluation of digital
printers in terms of their potential impact on print quality, in accordance with the quality
requirements of the NGAS.
The Color Research Laboratory received test prints from a total of six participants to
evaluate their print quality. To avoid any prejudices throughout the evaluation process
NGAS anonymised the test prints by identifying the competitors with A, B…F.
Firstly the results of the three categories individually will be presented before the final
ranking will be shown. Table 5 presents the results according to the visual evaluation of
the six test prints. Considering the “visual logo assessments” the experimental results
identify competitor A as the candidate that scored best in all the quality criteria in this
category.
Table 5: Results of the category visual logo assessments of six test prints.
Part 1: Visual logo assessments
Quality factors Weight A B C D E F
Colour match 50% 5.3 4.5 4.5 3.5 3.3 3.5
Visual resolution 20% 5 4.8 4.8 3.8 1.8 3.8
Surface texture 10% 4.8 4.8 3.8 3.8 3.3 2.8
Logo alignment 10% 5 5 4 4 3 5
Artefacts 10% 6 6 6 6 6 6
Total 26 25 23 21 17 21
Weighted mean 5.2 4.8 4.6 3.9 3.2 3.9
Ranking 1 2 3 4 6 4
150
Paper A
Especially for the quality factor “colour match”, which the NGAS has prioritised highest,
competitor A showed excellent performance. By presenting the results of the other
competitors, only B and C could match the minimal requirements of the expert panel.
Although the competitors D, E and F have shown reasonable results for the quality factors
«surface texture», «register» and «artefacts» the performance regarding «colour match»
was rather poor. In fact, the expert panel has decided that competitor D, E and F will be
excluded in the further evaluation process because they couldn’t match the requirements of
this category at all.
Table 6 presents the results of the second category, namely, the evaluation of quality
factors, which are not direct related to the colour logo. Nevertheless these quality factors
contribute to the entire image quality and may affect the rendering of the colour logo.
Table 6: Results of the category general print quality assessments of the

competitors A, B and C.
Part 2: General print quality assessments
Quality factors Weight A B C
Text quality 25% 3 5 4
Colour gamut 25% 6 4 3
Repeatability 25% 1 6 5
Register 25% 6 6 5
Total 16 21 17
Weighted mean 4 5.3 4.3
Ranking 3 1 2
It can be seen, that competitor B performs best in this category followed by competitors C
and A. Due to the imperfect repeatability performance obtained by competitor A the total
score within this category is rather poor. However, considering the quality factor “colour
gamut” competitor A has shown the largest range of colours achievable compared to the
other two competitors.
151
Paper A
Table 7: Results of the category copy quality assessments of the

Part 3: Copy quality assessments
Quality factors Weight A B C
Colour match between original print and copy 40% 5 2.8 4
Copy accuracy 40% 3.3 3.8 3.3
Artefacts 20% 4.3 4.5 4
Total 13 11 11
Weighted mean 4.2 3.5 3.7
Ranking 1 3 2
Table 7 presents the results of the category copy quality assessment according to visual
evaluation. Considering the quality factor “colour match between original and copy”
competitor A performs best, whereas competitor B shows some obvious limitations.
Although competitor B performs best regarding the quality factor “copy accuracy” all three
competitors achieve a reasonable result.
Table 8: Final evaluation results including the three categories and their
corresponding weights from competitor A, B and C.
Final ranking: part 1, part 2 and part 3 Weight A B C
Weighted mean part 1 50% 5.2 4.8 4.6
Weighted mean part 2 30% 4 5.3 4.3
Weighted mean part 3 20% 4.2 3.5 3.7
Total 8.2 8.8 8

Weighted mean 4.6 4.7 4.3
Final ranking 2 1 3
To obtain the final ranking the results of all three categories has been weighted according
to the priorities of the quality requirements defined by NGAS. Although Table 8 shows an
insignificant difference between competitor A and competitor B in terms of weighted
mean, competitor B has scored best considering the final ranking.
152
Paper A
4 Discussions
Although the results of our evaluation presented in the previous section are interesting, and
the methods are valid themselves, it is appropriate to proceed to a critical discussion.
Even if all three competitors showed an acceptable result in the category “general print
quality assessment” competitor B demonstrated the overall best performance in this
category. Considering the quality factor “text quality” all competitors have shown an
acceptable performance in terms of legibility for the font size of 3 pica. However,
considering the quality factor “colour gamut” Figure 5 shows that competitor A represents
the largest range of colours achievable on this substrate whereas competitor C has a
significantly smaller colour gamut.
In terms of matching the Pantone colours of the national colour logo the results
from competitor A show a strong correlation between the results of the quality factor
«colour match» in the category “visual logo assessments” and the quality factor “colour
gamut” in the category “general print quality assessments”.
Colour match of the logo colours

∆E*ab
18.00
16.00
14.00
12.00
10.00
8.00
6.00
4.00
2.00
0.00
Pantone 280M Pantone 185M Pantone 116M
Competitor C 10.14 15.60 14.47
Competitor B 12.30 15.91 5.43
Competitor A 4.14 11.76 13.99
Figure 4: Colour difference in terms of mean error between the Pantone colours and the
test print.
Considering the colorimetric measurements of the test print comparing with the target
colours Pantone 280M, Pantone 185M and Pantone 116M, non of the three competitors are
able to perform an absolute colour match, as seen in Figure 4. However, for the target
colour Pantone 280M competitor A shows a colour difference of ΔE*ab < 5 units whereas
competitor B and competitor C show a colour difference of ΔE*ab > 10. Regarding the
target colour Pantone 185M it can be seen that all three competitors have their limitations
in terms of colour match accuracy.
153
Paper A
Figure 5: Colour gamut projection of the competitors A, B and C onto the a*b*-plane
according to the level of lightness (L*) of the three colours of the national logo.
Figure 5 shows the colour gamut projection of the three competitors onto the a*b*-plane
according to the level of lightness (L*) of the three “optimized logo colours” (Table 3).
Furthermore it illustrates the distance from the nearest possible point in the colour space
achieved from the print medium and the target colour. Except for Pantone 116M,
competitor A shows the best performance in terms of colour difference between the
Pantone colours and the test print. The Pantone 185M was in general the most difficult
colour to reproduce.
Table 9: Size of colour gamuts, quantified as the volume of the convex hull of the gamut in
CIELAB colour space derived to the test prints of the three competitors A, B and C.
Competitor: CIELAB volume Relative volume
A 385000 unit 100%
B 295000 unit 77%
C 220000 unit 57%
Another way of looking at the size of colour gamuts is by considering the gamut volume.
The size of colour gamuts, quantified as the volume of the convex hull of the gamut in
CIELAB colour space was derived from the test prints of the three competitors A, B and
C. It can be seen in Table 9 that competitor A provides the largest gamut volume, whereas
the volume given by competitor C is only 57% of that given by competitor A.
154
Paper A
Repeatability
∆E*ab
6.00
5.00
4.00
3.00
2.00
1.00
0.00 Time
1 2 3 4 5 6 7 8 9 10
Competitor A 3.13 5.50 2.97 3.48 3.86 4.17 3.03 3.59 3.95 5.25
Competitor B 1.15 1.06 0.99 0.75 1.06 1.14 1.00 1.42 0.98 1.16
Competitor C 1.59 2.02 2.60 1.46 2.14 1.70 1.44 1.68 1.22 2.59
Figure 6: Repeatability variations over time in terms of ΔE*ab of

the three competitors A, B and C.
To quantify the effect of the printer’s repeatability, a digital test chart including 288 colour
patches, which representing the achievable colour gamut on a certain substrate, was printed
ten times at short-time interval. Figure 6 shows the repeatability variations according to ten
measurements in sequence. Apparently competitor B performs best compare to the other
two competitors with very small variations. Competitor C presents a larger variations but
the result is still within an acceptable tolerance. However it is worth pointing out that
competitor A has shown an unexpectedly poor repeatability performance. Currently the
exact cause of this could not be determined, although the most likely cause is human error
of the operator.
Finally, in Table 7, the results of the category “copy quality assessments” indicate the
best characteristics for competitor A in terms of “colour match between original print and
copy”. Nevertheless all three candidates have performed acceptably in terms of the quality
factors “copy accuracy” and “artefacts»”.
155
Paper A
Colour difference between original and copy

∆E*ab
40.00
35.00
30.00
25.00
20.00
15.00
10.00
5.00
0.00
Mean Standard deviation Max
Competitor A 11.93 7.08 34.26
Competitor B 11.68 5.79 29.11
Competitor C 8.20 3.88 22.86
Figure 7: Colour difference between original and copy in terms of ΔE*ab of the three
Although the final ranking list shows an insignificant difference between candidate A and
candidate B, competitor B has demonstrated the best performance. It is obvious that
competitor A would have achieved the overall best result by performing an acceptable
repeatability performance.
As seen in Table 7, competitor A has performed best regarding the visual evaluation of
the copy test. However the results of the colorimetric measurements between the original
and the copy do not match the visual evaluation in terms of colour difference. Figure 7
illustrates that competitor A shows the highest value according to the ΔE*ab colour
difference (Mean ΔE*ab 11.9 units) whereas competitor C shows the best performance. The
fact that there is no strong correlation between the visual evaluation and the colorimetric
measurement might be due to the different number of colours, which have been used for
the two methods of quality assessment. For the visual evaluation, the three logo colours
only have been considered whereas for the colorimetric measurement a large number of
colours which represent the entire reproducible colour space have been the target.
Suppose the weight of the chosen categories and defined quality factors had been
changed, the final result may have been different. Moreover it is worth pointing out that we
have proposed further quality factors, which could have been applied to this study to
evaluate the print quality (e.g. visual image quality, colorimetric reproduction, uniformity
and addressability). However, commercial and political circumstances of NGAS have
affected the consideration of the applied factors in this study.
156
Paper A
6 Conclusions and perspectives
The evaluation method including the selection of the quality factors used in this particular
project has been unique in terms of print quality. Although various factors have been
chosen and applied in the evaluation process the main target to assess the competitors has
been the accuracy of the colour reproduction of the three colours of the national logo. The
outcomes in this study demonstrate a strong correlation between the performances of the
competitors in terms of the obtained results of the various quality factors. The competitors,
that have shown good results considering the size of the colour gamut and the accuracy of
the colour match, have obtained reasonably good results in other quality factors too.
Except for one competitor, which has shown a very poor performance regarding
repeatability. However, the result of the printer evaluation can be used by the NGAS to
determine the final purchase decision in terms of image quality.
A similar approach might be adopted for other printer evaluation to justify the
appropriate equipment to match the desired print quality and requirements. The proposed
categories and quality factors are adjustable according to the quality requirements. After
the NGAS has decided which type of equipment shall be purchased and implemented
further measurement can be conducted on this new machine to confirm the results from the
prior tests. Nevertheless, the similarity between the purchased printers themselves in terms
of colour accuracy and stability over time would be a possibility for further studies.
There might be several potential directions for further work on this topic. The
acceptability threshold considering print quality assessment is defined as a vague concept
and one that depends strongly on application and industry. Moreover, a further quality
metric, which could be considered more suitable for the printer market is a tolerance
threshold.
Acknowledgments
The authors thank NGAS for their kind permission to use this material. Furthermore the
authors wish to extend their thanks to the participating members of the NGAS for their
grateful collaboration.
References
Farup I., Hardeberg J. Y., Bakke A. M., Kopperud S. and Rindal A. (2002). Visualization
and interactive manipulation of color gamuts. In Proceedings of IS&T and SID’s 10th
157
Paper A
Color Imaging Conference: Color Science and Engineering: Systems, Technologies,

Applications, pages 250–255, Scottsdale, Arizona.
Field G. (1999). Color and Its Reproduction. Second Edition, GATFPress, Pitttsburgh US
Gescheider G. A. (1985). Psychophysics, Method, Theory, and Application, Second

Edition, Lawrence Erlbaum Associates, ISBN 0–89859–375–1, 152.
Handley J. (2001). Comparative Analysis of Bradley-Terry and Thurstone-Mosteller

Paired Comparison Models for Image Quality Assessment, PICS 2001, 108-112.
Hardeberg J. Y. and Skarsbø S. E. (2002). Comparing color images quality of four digital
presses. Proceedings of the 11th International Printing and Graphic Arts Conference,
Bordeaux, France.
Ifra (2002). Colour quality audit in reproduction and print, Ifra Consulting Module
ISO 12647-1 (1996). Graphic technology - Process control for the manufacture of half-tone
colour separations, proof and production prints – Part 1: Parameters and measurement
methods, First edition
ISO 12647-2 (1996). Graphic technology - Process control for the manufacture of half-tone
colour separations, proof and production prints – Part 2: Offset lithographic processes,
First edition
ISO 3664 (2000). Viewing conditions – Graphic technology and photography – Second
edition
Kulturdepartementet (2000). Grafisk designprogram for departementene, Håndbok, annen

utgave, Statens forvaltningstjeneste
158
Paper A
Jacobson R. E. (1995). An Evaluation of Image Quality Metrics, The Journal of

Photographic Science, Vol.43 1995
Marcu G. (2000). Color Quality in Desktop Printing. Tutorial Notes, IS&T’s PICS
Conference, Portland, Oregon.
Microsoft Corporation (2001). Windows Color Quality Specifications for Printer OEMs.
Part of Microsoft Hardware Quality Labs (WHQL)’s Windows Color Quality Test Kit.
http://www.microsoft.com/hwdev/tech/color/ColorTest.asp.
Morovic J. and Nussbaum P. (2003). Factors Affecting The Appearance Of Print On

Opaque and Transparent Substrates, Journal of Imaging Science and Technology, JIST 47
#6
Stokes, M. (1998). The impact of color management terminology on image quality. In

Proc. IS&T’s 1998 PICS Conference, pages 174–178, Portland, Oregon.
159
Paper A
160
Paper B
Paper B
Peter Nussbaum and Jon Yngve Hardeberg
Print Quality Evaluation and

Applied Colour Management in Heat-set Web Offset
Published:
In Advances in Printing and Media Technology:
Proceedings of the 33rd International Research Conference of iarigai,
Volume 33, pp. 399-411,
N. Enlund and M. Lovreček, Ed.,
Acta Graphica Publishers,
2006.
161
Paper B
162
Paper B
Print Quality Evaluation and Applied Colour Management

in Heat-set Web Offset
Peter Nussbaum and Jon Y. Hardeberg
The Norwegian Color Research Laboratory,

Faculty of Computer Science and Media Technology,
Gjøvik University College, P.O.Box 191, N-2802 Gjøvik, Norway
E-mail: peter.nussbaum@hig.no, jon.hardeberg@hig.no
Keywords: Process control, colour measurement, print quality assessment, psychophysical

methods, colour management
Abstract
This paper aims to investigate print quality in heat-set web offset by applying colour
management. In particular it looks at the colorimetric properties of five heat-set web offset
presses in order to evaluate the appropriate colour separation approach, either by applying
individual separation profiles or by using industry standard profile such as
ISOwebcoated.icc. The key method underlying the work described here relies on obtaining
colour measurements to determine the repeatability of each participant in terms of colour
differences. Furthermore the variation between the five heat-set web offset printing
processes and the variation according to the colorimetric values of the ICC profile
“ISOwebcoated” are important parts of the quantitative evaluation. According to the colour
measurements two custom ICC profiles were generated and applied to four test images,
which were printed of the five heat-set web offset presses in a second test run. Furthermore
the industry standard ICC profile “ISOwebcoated” was applied too. A psychophysical
experiment was carried out to determine naturalness of the reproductions made according
to the three profiles applied. Finally the results of the study indicate the performance of the
appropriate profile applying to the five heat-set web offset presses to obtain significant best
print quality.
163
Paper B
1. Introduction
Although process control for the production of half-tone colour separations, proof and
production prints are clearly defined in ISO 12647-2:2004 often printing processes show
major variations, which affect the appearance of print. Essentially, there are two different
approaches considering printing press convention, namely optimized or standardized press
behaviour. A fully optimized press is all about maximizing its capability in terms of lowest
possible dot gain, highest ink densities and best contrast the individual printing press can
achieve without regards to any external specifications or standards. Such individual
parameters can create unique press condition, which requires custom ICC profiles to create
the appropriate separations. Another approach is to make the press conform to a certain
reference or standard such as ISO 12647-2:2004. By using the second approach and by
standardising the behaviour of the press industry standard ICC profiles can be used.
Recently, five of the largest Norwegian heat-set web offset printing plants started a
collaboration to evaluate their common print quality and print control to strengthen their
position in the heat-set market. Due to the print-on-demand concept, parts of a total
publication edition can be printed in different print locations. However, the appearance of
the total print edition must be identical. Therefore the variations of the five heat-set web
offset printing plants had to be verified to determine the appropriate adjustments. The
project is named “webicc” and the five participants are AllerTrykk, Hjemmet Mortensen
Trykkeri (HMT), Aktietrykkeriet, Norprint and Ålgård.
The aim of the presented work is to evaluate five heat-set web offset printing presses in
terms of their conformance to specified values, in accordance with the requirements of ISO
12647-2:2004. Furthermore, the assessment of each individual printing press and the
variation within the five participants are important parts of this study, in order to evaluate
the appropriate colour separation approach, either by applying individual separation
profiles or by using industry standard profile such as ISOwebcoated.icc. In order to obtain
the defined goals, the “webicc-group” in collaboration with the Norwegian Color Research
Laboratory has carried out this print quality project.
The next section of this paper will introduce the experimental method used for evaluating
the print quality. Then the data analysis performed on the experimentally obtained data
will be described, followed by a presentation of the results. Finally the implications of the
results will be discussed and ideas will be suggested for future work.
164
Paper B
2. Research Method
has been concluded as a very complex issue. The subject has been discussed at various
conferences (see e.g. Hardeberg and Skarsbø, 2002; Nussbaum et al., 2004).
Principally there are two different approaches to assess image quality. The first approach is
by measurement, using instruments to determine values for the various quality factors. The
second method is based on observation, using psychophysical experiments to gather the
judgment of human observers. For instance, the pair comparison method is a robust
approach where observers are asked to compare the perceptual magnitude between two
stimuli (or pairs of stimuli). This method is based on Thurstone’s «law of comparative
judgment» (Handley, 2001; Gescheider, 1985).
Considering the print quality evaluation in this study both methods have been applied,
using quantitative analyses based on colour measurements and psychophysical experiments
(Figure 1).
Figure 1 Overview of method
2.1 Quantitative evaluation

The purpose of the following test is to determine the print run repeatability of each
participant of the “webicc” project. Furthermore the variation between the five heat-set
web offset printing processes and the variation according to the colorimetric values of the
ICC profile “ISOwebcoated” are important parts of the quantitative evaluation. Hence, the
characterization test target ECI2002 was used in this work. As seen in Figure 1 two test
print runs with four months interval were carried out in this project. The colour
165
Paper B
measurement data from the first test print run were the source to generate two custom
printer profiles, which were applied in the second test print run.
Considering the substrate, a paper manufacture supplied all the participants with the
appropriate paper, type 3, 70g/m2. The ink set colours used in the test was supplied by two
manufacturers according to ISO 12647-2:2004. The colour measuring conditions were
according to ISO 13655:2000, geometry 45/0, 2ºobserver, CIELAB system. For
documentation all measuring were made on white backing. Considering colour difference
calculations ΔE*ab values were computed between individual measurements and the
“ISOwebcoated” colorimetric values. The arithmetic mean, standard deviation and
maximum of the resulting colour difference distributions were then computed.
The quantitative evaluation has been achieved on the basis of data gathered in an
experiment involving colour measurement. The Short-term repeatability performance of
the measuring instrument (GretagMacbeth Spectrolino) has a mean ΔE*ab of 0,02 (10
measurements at 10 sec. interval on white). Hence the instrument can be considered to
have a high degree of repeatability. The mean repeatability error of the reference
measuring set-up was ΔE*ab of 0,17 (equal test print measured two times).
2.2 Psychophysical Experiment

The aim of the psychophysical experiment was to determine naturalness of the
reproductions made according to different web coated prints. As a result of the colour
measurements of the first test run two ICC profiles were generated, namely, “webicc” and
“custom”. Consequently the psychophysical experiment was carried out according to the
test prints of the second test print run. In addition to the two custom profiles the
“ISOwebcoated” ICC profile was applied to each test image and printed in the five heat-set
web offset printing plants. The properties of the three applied ICC printer profiles are:
• ISOwebcoated: Based on the offset characterization table “FOGRA28L.txt” valid for

the following reference printing conditions relating to the international
standard ISO/DIS 12647-2:2003
• Webicc: ICC profile according to the average measurement values of all five
heat-set web offset printing plants (Profiling tool: GretagMacbeth,
ProfileMaker 5.0.4, predefined separation settings “Offset”)
• Custom: ICC profile (5 individual profiles) according to individual
measurement values of each heat-set web offset printing plant
166
Paper B
(Profiling tool: GretagMacbeth, ProfileMaker 5.0.4, predefined

separation settings “Offset”)
The source profile of the test images was “AdobeRGB(1998)”. Considering the conversion
options the Adobe (ACE) CMM and the relative colorimetric rendering intent has been
used. Four test images were chosen to contain a range of different types of pictorial content
and tonal and chromatic variety as shown in Figure 2. The chosen images represent typical
the type of images used in weekly magazines.
Figure 2 Four test images: camera, couple, portrait and house. The images “camera” and
“portrait” are reproduced with the permission of ” Ole Jakob Bøe Skattum”. The image
“couple” is reproduced with the permission of “Se og Hør” and the image “house” with
the permission of “Hjemmet Mortesen AS – Hytteliv”.
The images were simultaneously viewed side by side, as shown in Figure 3 and no anchor
stimuli were used as a reference, The viewing set up was based on the standard condition
of the graphic art industry ISO, 3664, vertical geometry 45°, background grey, light source
D50 simulator, light intensity 100%. The images were viewed in a darkened room (lights
off) and the images, approximately 10cm x 15cm in size, were viewed from a distance of
approximately 65 cm.
Figure 3 Viewing arrangement for psychophysical experiment
167
Paper B
A total of 27 observers (16 experts and 11 naïve) with normal colour vision and ages from
20 to 64 took part in the experiment. Each observer was required to make 3 pair-wise
comparisons per image (4) and per web offset process (5), a total of 60 comparisons. Note
that the comparisons were carried within each individual web offset press. The law of
comparative judgement was applied and the method used was pair comparison judgement
(Engeldrum, 2000). The observer’s task was to decide which of the two prints in the
viewing cabinet was most preferred in terms of naturalness. In a study by Morovic (2002)
naturalness is defined as an attribute, which is determined by identifying the locations of
prototypical memory colours such as “skin”, “grass” and “sky” in colour space (which
represent a wider range of naturally occurring colours) and then assessing image colours in
relation to these.
3. Results
As mentioned previously the aim of this work is to evaluate five heat-set web offset
printing presses in terms of their conformance to specified values, in accordance with the
requirements of ISO 12647-2:2004. Furthermore, the assessment of each individual
printing press and the variation within the five participants are important tasks to evaluate
the appropriate colour separation approach, either by applying individual separation
profiles or by using an industry standard profile such as “ISOwebcoated.icc”.
3.1 Results of the quantitative evaluation

The Color Research Laboratory received test prints from all five “webicc” participants and
measured a total of six (3 recto printing and 3 verso printing) test charts (ECI2002) per test
print run from each participant to evaluate their colorimetric properties.
To determine the variation in terms of colour stability over time the quality factor
repeatability is required (Morovic and Nussbaum, 2003). Figure 4 shows the performance
of the short-term repeatability of each participant. According to the first test print run the
colour differences were calculated between the average of the six measurements and each
individual measurement. Apparently, participant “Norprint” shows the best performance
with a mean ΔE*ab 1.2 units. Although the other four participants present larger variations
the results are still within an acceptable tolerance (Mean ΔE*ab between 1.5 and 2.0 units).
168
Paper B
2.5
2
Eab (mean)
AllerTrykk
1.5
HMT
Aktiettrykkeriet.
1 Ålgård
Norpint
0.5
0
Recto 1 Recto 10 Recto 21 Verso 1 Verso 10 Verso 21
Average Average Average Average Average Average

print number
Figure 4 Short-term repeatability performance of each participant
To determine the long-term repeatability performance from each participant colour

measurement data from the first test print run and colour measurement data from the
second test print run has been compared. As mentioned before the two test print runs were
carried with four months interval. As expected the results in Table 1 show higher
variations in terms of colour differences. However, as already seen in the short-term
repeatability participant “Norprint” shows again the best performance in the long-term
repeatability with a mean ΔE*ab 1.92 units. On the other hand participant “HMT” presents
the largest variation with a mean ΔE*ab 6.48 units.
Table 1 Long-term repeatability performance of each participant

AllerTrykk HMT Aktietrykkeriet Norprint Ålgård
Mean 2.68 6.48 3.92 1.92 4.53
STDEV 1.67 3.68 1.79 1.3 1.78
MAX 10.85 19.03 12.36 7.08 9.81
It is important to point out that the first test print run has been carried out according to
individual “in-house” density standards defined by each participant in the “webicc”
project. These density values are unknown and not available to publish. For the second test
print run the “webicc” group has defined a “webicc” density standard, which is required to
obtain by each participant. The “webicc” densities of dry solids were measured without
polarizing filter. The values are for black 1.4, cyan 1.25, magenta 1.2 and yellow 1.1.
Another approach to assess the colorimetric properties is to analyse the variations between
the five heat-set web offset printing processes in the first test print run. Figure 5 presents
the colour difference between the average measurements of all five heat-set web offset
169
Paper B
plants and each individual participant. Notice, that participant “HMT” shows the largest
difference (mean ΔE*ab > 4 units) and participant “Norprint” illustrates the smallest
difference.
14
12
10
8
Eab
0
Mean STDEV MAX
Figure 5 Variations between the five participants
An important part of the evaluation is to analyse the variation between all five “webicc”
participants and the colorimetric values of the “ISOwebcoated” ICC profile. According to
the results given in Table 2, the participant “AllerTrykk” shows the best match (mean
ΔE*ab < 5 units) and participant “HMT” presents the largest difference (mean ΔE*ab > 8
units). Note that all the measurements from the first test run were carried out according to
individual “in-house” density standards. Although paper type 3 was used in the project
minor variations in terms of colour differences (Table 3) according to ISO 12647-2:2004
and “ISOwebcoated” were detected which also may contribute to the overall colour
differences.
Table 2 Colorimetric colour differences between each heat-set web offset plants and
ISOwebcoated values
ISOwebcoated ISOwebcoated ISOwebcoated ISOwebcoated ISOwebcoated
Mean 4.97 8.15 7.09 6.39 6.69
STDEV 2.64 3.97 3.64 3.51 3.26
MAX 13.76 21.11 17.55 16.67 18.26
Table 3 CIELAB coordinates for paper type 3

L* a* b*
ISOwebcoated 92.4 -0.7 1.5
webicc 90.2 0.3 -0.7
170
Paper B
ISO 12647-2:2004 92 0 5
Tolerance ±3 ±2 ±2
Another way of looking at the colour differences between colour measurement data of each
participant and the data of the “ISOwebcoated” profile is to consider the colour patches
with the highest fluctuations. Figure 6 shows the colorimetric colour differences between
each heat-set web offset plant and the “ISOwebcoated” values including the worst 10%
colour differences of the colour patches which are outlined in yellow.
Figure 6 The worst 10% colour differences outlined in yellow between each heat-set web
offset plant and “ISOwebcoated”
Looking at the CIELAB values for the primary colours cyan, magenta, yellow and black
specified in ISO 12647-2:2004 the differences between the actual values and the nominal
values must not exceed the tolerances shown in Table 5. Table 4 presents the CIELAB
values for the primary colours from all five “webicc” participants, “ISOwebcoated” and
the ISO 12647-2:2004 specification. Moreover the Table shows colour differences
calculated between the participants, “ISOwebcoated” and the ISO 12647-2:2004 from the
first test print run (1.TP) and the second test print run (2.TP). As expected
“ISOwebcoated” is within the tolerance values ΔEab < 5 units for all four primary colours.
For the primary colours cyan and black all five “webicc” participants show reasonable
results in terms of tolerances. Only participant “HMT” shows colour differences above the
CIELAB tolerances. On the other hand the performance of all five participants is rather
171
Paper B
poor for the primary colours magenta and yellow comparing to the “ISOwebcoated”
CIELAB values. Note that the target value b* for the primary colour yellow in ISO 12647-
2 is defined by b*=94. Apparently, none of the participants is able to obtain an appropriate
match. Consequently the colour differences in yellow are rather high for all participants.
This has also been confirmed, in Figure 6 where the worst 10% colour differences of the
colour patches are outlined in yellow, which are typically in the area of yellow, magenta
and red.
Table 4 CIELAB coordinates for paper type 3 according to measurements on white
backing (ISO 12647-2:2004) for 1. Test Print (1.TP) and 2. Test Print (2.TP)
Cyan Magenta Yellow Black
1.TP 2.TP 1.TP 2.TP 1.TP 2.TP 1.TP 2.TP
House webicc House webicc House webicc House webicc
density density density density dens density density density
AllerTrykk L* 57.5 -55.1 49.2 49.1 83.9 84.5 22.1 21.2
a* -35.4 -36.1 68.9 69.7 -3 -3 1.8 1.5
b* -41.2 -43.8 -8.8 -7.2 83.1 85.2 3.2 2.8
ΔE*ab 4.6 1.9 10.7 8.9 12.1 9.9 4.2 3.4
HMT L* 50.6 55.6 46.6 50.9 82.9 84.3 13.8 25.1
a* -35.9 -35.4 72.1 66.1 -0.6 -2.4 0.7 2
b* -46.6 -43.3 -2.6 -8.6 81.5 82.1 0.7 4.5
ΔE*ab 5.5 2.8 4.6 13 14.3 13 6.3 7.1
Aktietrykk
L* 56 60.3 50.6 50.6 83.8 84.7 23.2 22.9
eriet
a* -34.8 -33.7 66.3 67.1 -2.5 -2.8 1.1 1
b* -42.7 -39.7 -9.5 -7.8 82.1 77.5 2.5 0.9
ΔE*ab 3.6 8.1 13 11 13.1 17 4.2 3.2
Norprint L* 56.9 56.2 47.6 47.8 84.4 83.6 18.9 20.4
a* -37.8 -37.6 70.1 70 -3 -2.1 0.1 0.4
b* -42.7 -43 -6.5 -5.8 76.7 79.2 1.8 2
ΔE*ab 2.3 1.6 8.3 7.8 18.0 16 2.1 2.1
Ålgård L* 55.8 57.6 48.9 49.5 84 85 21.8 22.6
a* -37.2 -36.1 68.4 67.2 -2.4 -3.3 1.6 1.7
b* -44 -41.5 -7.4 -8.3 80.4 75.3 2.9 1.7
ΔE*ab 1.1 4.1 9.9 11 14.6 19 3.8 3.5
ISO
L* 55.5 46.2 85.8 17.6
webcoated
a* -38.6 71.2 -2.9 0.6
b* -44.7 -1.7 93.3 0.3
ΔE*ab 1.0 5.0 3.5 2.5
ISO
L* 55 49 89 20
12647-2
a* -38 75 -4 0
b* -44 0 94 0
< CIELAB tolerances > CIELAB tolerances
172
Paper B
Table 5 CIELAB tolerances (ΔE*ab) for the solid tones of primaries according to ISO
12647-2:2004
Deviation 5 5 5 5
Figure 7 shows the colour gamut projection of the five heat-set web offset presses and the
ICC profile “ISOwebcoated” onto the a*b*-plane according to the level of lightness
L*=50.
The data used in this task are based on colorimetric measurements, which have been
analysed and visualised by the icc3d application (Farup et al., 2002). The Figure to the left
(Figure 7A) illustrate the gamut projection of the five heat-set web offset presses according
to the first test print and the Figure to the right (Figure 7B) presents the gamut projection
of the second test print with “webicc” density. It can be seen that the “ISOwebcoated”
profile has a slightly larger colour gamut than the five participants of the “webicc” project.
ISOwebcoated AllerTrykk HMT Aktietrykkeriet Norprint Ålgård

Figure 7 2D colour gamuts comparison of the five heat-set web offset presses and the ICC
profile ISOwebcoated onto the a*b*-plane according to the level of lightness L*=50
A) First test print run according to “in-house” densities B) Second test print run
according to “webicc” density
Note that the size of the colour gamut of the five participants has changed from the first
test print to the second test print. Remember that the density values from the first test print
run are not identical to the density of the second test print. Furthermore variations in terms
of long-term repeatability affect the colour gamut too. However, as seen in Figure 7B the
improvement of the “webicc” density values results in a closer match to the
173
Paper B
“ISOwebcoated” colour gamut. Particularly participant “HMT” shows in the first test print
(Figure 7A) a much larger gamut in the green blue area. Further the largest colour
differences between “ISOwebcoated” and the “wecicc” participants observed previously in
the primary colours magenta and yellow (Table 4) can also be seen in the colour gamut
comparison in Figure 7.
The three-dimensional CIELAB plot in Figure 8 reveal the colour gamut properties of the
“ISOwebcoated” profile and the “webicc” profile which contains the average measurement
values of all five heat-set web offset printing plants. The plot shows obvious limitations of
the webicc profile in high-saturated colours such as yellow and magenta. While the
“ISOwebcoated” profile has some minor limitation in the blue area comparing to the
“webicc” profile.
Figure 8 3-D comparison of the “ISOwebcoated” colour gamut (wireframe red) and the
“webicc” colour gamut (wireframe green)
Another way of looking at the size of colour gamuts is by considering the gamut volume.
The size of colour gamuts, quantified as the volume of the convex hull (Morovic, 2003) of
the gamut in CIELAB colour space was derived from the test prints of both test print runs
of the five heat-set web offset presses and the ICC profile ISOwebcoated.
Table 6 Comparison of the approximate relative gamut volumes of the five heat-set web
offset presses and the ICC profile “ISOwebcoated”
Participants: 1. Test Print 2. Test Print
House density Webicc density
CIELAB Relative CIELAB Relative
volume volume volume volume
ISOwebcoated 436’900 unit 100% 436’900 unit 100%
174
Paper B
AllerTrykk 432’646 unit 99% 472’158 unit 108%

HMT 482’392 unit 111% 437’106 unit 101%
Aktietrykkeriet 423’847 unit 97% 423’556 unit 97%
Norprint 415’451 unit 95% 422’646 unit 96%
Ålgård 428’605 unit 98% 382’246 unit 87%
The density changes from the first test print run (“house-density”) to the second (“webicc”
density) affect the gamut volume. It can be seen in Table 6 that AllerTrykk provides the
largest gamut volume (even 8% larger than the “ISOwebcoated”) in the second test print
run, whereas the volume given by participant “Ålgård” is 87% of that given by
“ISOwebcoated”. Although participant “HMT” has shown the largest mean colour
difference comparing to “ISOwebcoated” as seen in Table 2, the relative gamut volume is
virtually equal to that of the “ISOwebcoated”. This reinforces the fact that colour
differences say something more specific about individual colours whereas relative gamut
volumes refer to their ranges.
3.2 Results of the psychophysical experiment

The following are the results in terms of z-scores, which have been obtained in the
psychophysical experiment. For each image and heat-set web offset printing plant the 3 x 3
matrices of comparisons results for each observer were arranged over the 27 observers and
transformed into z-scores. The precisions of the experimental results are described in terms
of 95% confidence interval (CI), which is calculated using equation (1) using the mean (R),
standard deviation ( σ ) and the number of observations (N). Using case V of the method
proposed by Thurstone, the standard deviation of the z-scores is assumed to be σ = 1/ 2 .
σ
95% confidence interval = R ±1.96 (1)
N
175
Paper B
1.00
z-scores 0.50
0.00
-0.50
-1.00
-1.50
ISOWEBCOATED WEBICC CUSTOM
AllerTrykk -0.07 -1.14 0.65
HMT 0.61 0.51 -0.86
Aktietrykkeriet 0.36 0.58 -0.65
Algard 0.22 -0.30 -0.07
Norprint 0.72 -0.91 -0.26
Separations profiles
Figure 9 Pair comparison z-scores for each heat-set web press and separation (The error
bars represent 95% of population distribution).
Figure 9 presents the results in terms of z-scores for each heat-set web press. For the
number of images and observations, CI was calculated to be ±0.13. Except for the printing
plant “AllerTrykk” the profile “ISOwebcoated” shows the best performance for the other
four printing plants. The profile “custom” was ranked significantly best and the profile
“webicc” significantly worst for “AllerTrykk”. Note that the profiles “ISOwebcoated” and
“webicc” do not indicate a significant difference in terms of preferred naturalness. It is
interesting to note that despite a large colour difference in the quantitative evaluation
(Table 2) between the colorimetric values of the participant “HMT” and the profile
“ISOwebcoated” the performance with the ISOwebcoated profile ranked significantly best.
On the other hand considering the long-term variation between the first test print run and
the second test print run the participant “HMT” has shown the largest colour difference.
Hence the “custom” profile for “HMT” performs worst.
176
Paper B
1.00
0.80
0.60
0.40
z-scores
0.20
0.00
-0.20
-0.40
-0.60
-0.80
Camera 0.40 -0.08 -0.36
Couple 0.09 -0.32 0.20
House 0.44 -0.25 -0.32
Portrait 0.61 -0.36 -0.44
Figure 10 Pair comparison z-scores for each type of image and separation (The error bars
represent 95% of population distribution).
Considering Figure 10 the experimental results identify the profile “ISOwebcoated” as the
separation, which scored significant best for the images “Camera”, “House” and “Portrait”.
For the image “Couple” the profile “custom” performed best in terms of preferred
naturalness. Considering the images “Camera”, “House” and “portrait” the profile
“custom” shows a lower performance comparing to the profile “webicc”. However there is
no significant performance difference between the profile “webicc” and the profile
“custom”.
Table 7 Ranking of separation profiles for each image (1=best, 3=worst)
Image ISOWEBCOATED WEBICC CUSTOM
Camera 1 2 3
Couple 2 3 1
House 1 2 3
Girl 1 2 3
Overall 1 2 3
Table 7 gives the ranking of the performance of the three separation profiles for the four
images. As it can be seen the profile “ISOwebcoated” was significant best and the profile
“webicc” was second and the profile “custom” was third respectively. Figure 11 presents
the results in terms of z-scores for all four images and all five webicc-project participants.
For the total number of images, heat-set web offset presses and observations, CI was
calculated to be ±0.06. Notice, overall, the “ISOwebcoated” profile performed significant
177
Paper B
best. Although the results don’t show a significant difference between the “custom” and
the “webicc” profile the “custom” profile performed second best.
0.5
0.4
0.3
0.2
z-scores
0.1
-0.1
-0.2
-0.3
-0.4
Figure 11 Overall z-scores for the three separation profiles
4. Discussion
Considering the colorimetric results and the colour differences between the individual
printing plants and the colorimetric data of the “ISOwebcoated” profile (Table 3) it might
be expected that a custom profile will perform best. However, it is worth pointing out that
the characterization data for generating the “ISOwebcoated” profile were obtained by
measuring a large number of printed samples including averaging, rounding and further
adjustments to smoothen the data. Although the results of the repeatability of the
individual presses have shown reasonable performances small deviations in the print
process can cause unwanted variations in the appearance of the print by using custom
profiles. As seen in Figure 9 the “webicc” profile performs similar comparing to the
“custom” profile. Due to the fact that the “webicc” profile is related to a larger number of
measurements comparing to the “custom” profile the behaviour is similar to the
“ISOwebcoated” profile considering the smoothness, which tolerates larger variation than
the “custom” profile. Furthermore it must be mentioned that the profile “webicc” and the
profile “custom” were generated using GretagMacbeth ProfileMaker 5.0.4 whereas the
profile “ISOwebcoated” has been generated using Heidelberg Printopen 4.0.5.2. Although
the predefined separation settings “Offset” in ProfileMaker 5.0.4 has been chosen, which
178
Paper B
is, according to GretagMacbeth, compliant with the specifications laid down by FOGRA
and ISO 12647-2 for glossy and matt papers the separations using by “ISOwebcoated”
gives a results with a much stronger GCR (grey component replacement) degree. The test
images “camera” and “portrait” contain large areas with neutral colours, which might be
easier exposed to print variations and finally affect the appearance of the print. Hence the
reproductions made with the “ISOwebcoated” profile gives better print results in terms of
preserving grey balance. Note, based on the psychophysical experiment the results shown
in Figure 10 demonstrate a significant best performance for the test images “camera” and
“portrait” reproduced by the “ISOwebcoated” profile.
It is worth pointing out that in the beginning of the project a generic Norwegian heat-set
web offset profile has been considered for application in the Norwegian market. However,
the Norwegian Color Research Laboratory argued very soon to make the presses conform
to a certain reference or standard such as ISO 12647-2. As a result of the present study it is
possible for all participants of the “webicc”-group” to adopt the industry standard profile
“ISOwebcoated”. Nevertheless to preserve the daily printing conditions and to match the
colorimetric requirements of the adopted standard profile it is highly recommended press
control according to the standard ISO 12647-2.
5. Conclusions and Perspectives
As can be seen from the results of the psychophysical experiment the ICC profile
“ISOwebcoated” performs significant best than the other two profiles, “webicc” and
“custom” respectively. The poor performance of the “custom” and “webicc” profile is not
only due to the low degree of GCR. Rather a much larger number of printed samples
including averaging, rounding and further adjustments are required to smoothen the data.
Although some of the participants of the project still show significant colour variations
between their individual press behaviour and the colorimetric values of the ISO 12647-
2:2004 the use of the “ISOwebcoated” profile perform reasonable results. However, the
outcomes of the quantitative evaluation demonstrate clearly that there is still potential to
improve the target values of the ISO 12647-2:2004 to obtain a better coherence between
the “webicc” participants. Furthermore the outcomes in this study demonstrate an obvious
need for standardising the behaviour of the heat-set web offset presses.
Finally, it is of interest to consider other potential directions for further work in the field of
print quality assessment. Colour difference metrics for image quality assessment has been
179
Paper B
used widely for various applications. However there is not very often a strong correlation
between the objective evaluation and the visual assessment. Furthermore the interpretation
of the complete image quality assessment considering the colour difference calculation is
dependent on the application and the acceptance. The acceptability threshold considering
print quality assessment is defined as a vague concept and one that depends strongly on
application and industry. Moreover, a further quality metric, which could be considered
more suitable for the heat-set web offset printing process, is a tolerance threshold.
Acknowledgements
The authors thank “webicc” for their kind permission to use this material. Furthermore the
authors wish to extend their thanks to the participating members of the “webicc” for their
grateful collaboration.
Literature
Engeldrum P.G. (2000) Psychometric Scaling, A Toolkit for Imaging Systems
Development, Imcotek Press, Winchester USA, 1st ed.
Farup I., Hardeberg J. Y., Bakke A. M., Kopperud S. and Rindal A. (2002). Visualization
and interactive manipulation of color gamuts. In Proceedings of IS&T and SID’s 10th
Color Imaging Conference: Color Science and Engineering: Systems, Technologies,
Applications, pages 250–255, Scottsdale, Arizona.
Gescheider G. A. (1985) Psychophysics, Method, Theory, and Application, Second

Edition, Lawrence Erlbaum Associates, ISBN 0–89859–375–1, 152.
Handley J. (2001) Comparative Analysis of Bradley-Terry and Thurstone-Mosteller Paired

Comparison Models for Image Quality Assessment, PICS 2001, 108-112.
Hardeberg J. Y. and Skarsbø S. E. (2002) Comparing color images quality of four digital
presses. Proceedings of the 11th International Printing and Graphic Arts Conference,
Bordeaux, France.
ISO (2004) ISO 12647-2:2004 Graphic Technology – Process control for the production
of half-tone colour separation, proof and production prints.
180
Paper B
ISO (2000) ISO 13655:2000 Graphic Technology – Spectral measurements and

colorimetric computation for graphic arts images.
ISO (1999) ISO 3664:1999 Viewing conditions – Prints, transparencies and substrates for
graphic arts technology and photography.
Morovic J. (2003) Gamut mapping. In Digital Color Imaging Handbook. S. Sharma, CRC
Press, pp 639-685.
Morovic J. (2002) Colour gamut mapping. In Colour Engineering Achieving Device

Independent Colours, P. Green and L. MacDonald, John Wiley & Sons Ltd, pp 297-317.
Morovic J. and Nussbaum P. (2003). Factors Affecting The Appearance Of Print On

Opaque and Transparent Substrates, Journal of Imaging Science and Technology, JIST 47
#6.
Nussbaum P., Hardeberg J.Y. and Skarsbø S.E. (2004) Print quality evaluation for
governmental purchase decisions, In Proc IARIGAI Conference, Copenhagen, Denmark
Thurstone L.L. (1927) A law of comparative judgement. Psychol. Rev., 34, 273-286
181
Paper B
182
Paper C
Paper C
Aditya Sole, Peter Nussbaum and Jon Yngve Hardeberg
Implementing ISO 12646 Standards for soft Proofing in a Standardized

Printing Workflow according to PSO
Published:
In Advances in Printing and Media Technology:
Proceedings of the 37th International Research Conference of iarigai,
Volume 37, pp. 215-226,
N. Enlund and M. Lovreček, Ed.,
International Association of Research Organizations for the Information, Media and
Graphic Arts Industries,
2010.
183
Paper C
184
Paper C
Implementing ISO12646 Standards for soft Proofing in a Standardized

Printing Workflow according to PSO
Aditya Sole, Peter Nussbaum, Jon Y. Hardeberg

E-mail: aditya.sole@hig.no, peter.nussbaum@hig.no, jon.hardeberg@hig.no
Keywords: Colour measurement, colour management, process control standards, soft

proofing, display calibration, display characterisation
Abstract
This paper defines one of the many ways to setup a soft proofing workstation comprising
of a monitor display and viewing booth in a printing workflow as per the Function 4
requirements of PSO certification. Soft proofing requirements defined by ISO 12646 are
explained and are implemented in this paper.
Nec SpectraView LCD2180WG LED display along with Just colorCommunicator 2
viewing booth and X-rite EyeOne Pro spectrophotometer are used in this setup. Display
monitor colour gamut is checked for its ability to simulate the ISO standard printer profile
(ISOcoated_v2_300_eci.icc) as per the ISO 12646 requirements.
Methods and procedures to perform ambient light measurements and viewing booth
measurements using EyeOne Pro spectrophotometer are explained. Adobe Photoshop CS4
software is used to simulate the printer profile on to the monitor display, while, Nec
SpectraView Profiler software is used to calibrate and characterize the display and also to
perform ambient light and viewing booth measurements and adjustments.
1. Introduction
To reduce expensive and time consuming iterations in standardized printing workflow soft
proofing has become an important concept in predicting the final print product. Soft
proofing can be defined as ‘the ability to match colour images displayed on colour
monitors to the images produced when the same digital file is rendered by proofing and
printing systems’ (ISO12646 2008). However, although the concept soft proofing is not
185
Paper C
new and the task may sound rather simple, in practical applications the colour appearance
between two different medias (e.g. softcopy simulation of a hardcopy) can differ a lot due
to unsuitable type of devices, incorrect use of parameters and inaccurate device calibration
and characterization or inappropriate measurement methods. In the past a number of
studies and research work have addressed the issue of soft proofing. For more details see
the work of (Gatt et al. 2004) and (Gatt et al. 2005).
ISO 12646 defines parameters for monitor and viewing booth condition setup for soft
proofing environment. The practical methods to implement these standards as per the job
requirements are not been clearly defined. This paper therefore is aiming for to describe in
details one of the ways to setup an appropriate soft proofing station (comprising of a
monitor and a viewing booth with the appropriate ambient lighting conditions) according
to ISO 12646 standards for soft copy and hard copy proof comparison in the graphic arts
industry and evaluate the performance of the entire soft proof set up according to ISO
12646. This is also in accordance with soft proofing in a standardized printing workflow
according to PSO (UGRA 2009).
In this work the current standards and the appropriate parameters for soft proofing are
reported before proposing the method of implementing these standards.
2. Methods
2.1 Standards for soft proofing (specifications)

ISO 12646 specify the soft proofing standards. This standard describes two scenarios. The
first scenario consists of comparing a soft copy directly with a hard copy while the second
scenario consists of viewing soft copy images on a display independently of any hard copy
images. This paper primarily focuses on the first scenario where soft copy is directly
compared with the hard copy.
The technical requirements defined by ISO 12646 for scenario 1 i.e. comparing a soft copy
with a hard copy are as follows:
Ambient illumination, surroundings and environment

The level of ambient illumination should be sufficiently low. The standards recommend for
“luminance of a perfectly reflecting diffuser placed at position of the faceplate of the
monitor, with the monitor switched off, shall not be greater than ¼ of the monitor white
point luminance and should not be greater than 1/8 of the monitor white point luminance”
(ISO 12646, sec 4.7.2)
186
Paper C
The colour temperature of the ambient light, such as room light, should be within ± 200°k
of the colour temperature of the illumination used in the viewing booth. The conditions
within the viewing booth shall conform to viewing condition P2 of (ISO3664 2009)(ISO
12646, sec 4.7.2). P2 viewing conditions are defined as the conditions for practical
appraisal of prints, including routine inspection.
The luminance of the area surrounding the monitor shall not exceed 1/10 of the luminance
of the monitor showing a white screen. Extraneous light, whether from light source or
reflected by objects, shall be baffled from view and from illuminating the print or other
image being compared (ISO 12646, sec 4.7.2)
The surround and backing shall be neutral and matt (ISO 3664, sec 4.3.4). The surround
shall have a luminous reflectance between 10% and 60% with the specific value being
selected to be consistent with practical viewing.
Chromaticity, luminance of the white and black points

The black point of the display shall have a luminance that is less than 1% of the maximum
luminance of the display (ISO 12646, sec 4.8.1).
The conditions within the viewing booth shall conform to viewing condition P2 of ISO
3664 (ISO 12646, sec 4.8.2). P2 viewing conditions specify conditions applicable for the
appraisal of tone reproduction of individual images, photographic image inspection or the
judgement of prints. The illumination of the plane of viewing shall approximate that of
CIE standard illuminant D50. It shall have u’10, v’10 chromaticity co-ordinates within the
radius of 0.005 from u’10 = 0.2102, v’10 = 0.4889 in the CIE 1976 Uniform Chromaticity
Scale (UCS) diagram (calculations using the 10° observer angle). The illuminance at the
centre of the viewing surface shall be (500 ± 125 lux). The illumination uniformity should
be such that for a viewing area up to 1m2, the illuminance at any point within the
illuminated area shall not be less than 75% of the illumiance measured at the centre of the
illuminated viewing surface area. The uniformity should be evaluated by measuring at least
9 points equally distributed on the viewing surface (ISO 3664, sec 4.3.3).
The luminance of the white displayed on the monitor shall be at least 80 cd/m2 but
preferably 160 cd/m2 in order to match an unprinted sheet of white paper located close to
monitor having an illuminance of 500 lux, as specified in ISO 3664 for viewing condition
P2 (ISO 12646, sec 4.8.2).
ISO 12646 recommends D50 as white point for the soft proofing display (as the white
point in the viewing booth is D50); namely u’=0.2092, v’=0.4881, as specified in CIE
187
Paper C
Publication 15 (CIE15 2004). The chromaticity obtained, for the white point chosen shall
be within a circle of radius 0.005 from this point (ISO 12646, sec 4.8.2).
Colorimetric accuracy and grey balance of the display

Tristimulus values shall be measured for at least 10 neutral colours (R=G=B). For each
neutral colour the colour difference, ΔE*c = √(Δa2+Δb2), where Δa is the difference for the
CIELAB (red-green) co-ordinate and Δb is the difference for the CIELAB (yellow-blue)
co-ordinate, shall be calculated between these measured CIELAB values and the CIELAB
values which are intended to be displayed by the software characterizing the display. The
average deviation shall not exceed ΔE*c = 3 and preferably not ΔEc = 2 (ISO 12646, sec
4.10).
The average of the ΔE*ab between the measured CIELAB values and the CIELAB values
intended to be displayed by the *.icc monitor profile shall not exceed ΔE*ab 5 and
preferably not ΔE*ab 2 units respectively.
Uniformity of display luminance

The display should be visually uniform when displaying flat white, grey and black images.
All the measured luminance values should be within 5% of the luminance of the centre and
shall be within 10% of it (ISO 12646, sec 4.4).
Table 1 below summarizes the specification and tolerances defined by ISO 12646
standards for soft proofing.
Table 1 Specification and tolerances according to ISO 12646 standards for soft proofing
Parameters Target
Colour Intensity Chromaticity Uniformity Black point
Temp
Ambient 5000°K ¼ monitor Uniformly -NA-
light white point diffuse
Surround Neutral and matt with luminous reflectance between 10% and 60%
Display 5000°K >80 cd/m2 u’= 0.2092 10% within Luminance <
v’= 0.4881 the centre 1% of max
measurement luminance
2
Viewing 5000°K 500lux u’10 = 0.2102 for 1m area, -NA-
booth ± 125lux v’10 = 0.4889 luminance =
>75% luminance at the
centre
188
Paper C
2.2 Procedure followed for implementing the conditions/ specifications

In this section methods and procedures are proposed for implementing the specifications
mentioned previously. To setup the soft proofing station and the ambient light condition
following apparatus was used:
Hardware
• NEC SpectraView LCD 2180WG LED display
• Apple MAC PRO 10.4
• Just Normlicht ColorCommunicator 2 viewing booth
• X-Rite EyeOne Pro spectrophotometer including ambient light measurement
adaptor.
• OSRAM L58W/950, LUMLUX de LUXE, Daylight tubes for ambient light
Software
• SpectraView Profiler Version 4.1 (NEC SpectraView monitor calibration and
profiling software)
• UDACT (UGRA Display Analysis Certification Tool) monitor certifying
software by UGRA.
• Adobe Photoshop CS4
Nec SpectraView LCD 2180EG LED display was used as the display monitor to display
the soft copy images. Nec LCD 2180WG LED is a high end display which features
individual high power red, green and blue LEDs (Light Emitting diodes) as a backlight
light source for the LCD, instead of the typical CCFL (Cold Cathode Fluorescent Lamp)
(NEC 2005). LED backlight results in a wide output colour gamut of the display.
Hardware calibration can be performed on this display using its own calibration and
profiling software. Hardware calibration lets the user adjust the brightness and gradation
properties of each RGB primary within the monitor compared to the software calibration
were the monitor’s primaries are measured and the difference between the measured values
and the target values is corrected by adjusting the output of the video card. Software
calibration method can create problems like greyscale banding, decline in appearance of
colours in greyscale images. Hardware calibration eliminates the need to correct the RGB
output for smooth, accurate display of greyscale images.
189
Paper C
Nec SpectraView Profiler software calibrates and profiles the Nec LCD 2180WG LED
display monitor (NEC 2008). The software allows custom target calibrations that can be
preset or can be defined by the user. The user can define the white point, light intensity,
gamma curves, black point, etc. At the end it also profiles the display and generates a *.icc
colour profile. Target and measured results are analysed and displayed, showing
information like measured colour, colour difference, gamut etc. (NEC 2005).
Just colorCommunicator 2 viewing booth was used for the soft proofing station along with
the NEC SpectraView monitor display. Just colorCommunicator 2 can be connected to the
system by an USB interface. The light intensity of the viewing booth can therefore be
adjusted via software to match the light intensity of the display for soft proofing (JUST
2008). NEC SpectraView Profile software has a facility to communicate with the Just
colorCommunicator 2 to adjust the light intensity of the viewing booth as per the ISO
12646 soft proofing standards.
To perform the measurements of light intensity, white point, colour rendering on the
monitor display and on the hard proofing a number of different measuring instruments can
be used. To perform the measurement of colour on the monitor display either a colorimeter
or a spectrophotometer can be used. While to perform ambient light measurements a light
meter, colorimeter or a spectrophotometer with ambient light adaptor can be used.
To avoid inter-instrument uncertainty and reproducibility issue only one instrument has
been used for the measurements (Nussbaum et al. 2009). For this setup, a X-Rite EyeOne
Pro spectrophotometer was used to perform the measurements on monitor display and hard
copy and also to perform the ambient light measurements using the ambient light adaptor.
Monitor Calibration and viewing booth adjustment

To calibrate the monitor and adjust the viewing booth following procedure was followed:
The viewing booth and the monitor display were turned on half an hour before any
measurements/adjustments were made in order to properly warm and stabilize the
performance of the devices. The viewing booth was connected to the system via UBS cable
connection.
X-rite EyeOne Pro was connected to the system and the emission measurement mode
chosen. Consequently the LCD display profiling test chart (with 99 colour patches) was
measured using the ProfileMaker Pro MeasureTool software to warm up the measurement
instrument.
190
Paper C
NEC SpectraView Profiler Version 4.1 software was used to calibrate, to characterize and
to adjust the viewing booth according to ISO 12646 standards parameter for soft proofing.
This software communicates with the viewing booth via the USB cable connection (NEC
2008). P2 viewing conditions defined in ISO 3664 were setup in the viewing booth
according to ISO 12646 specifications (ISO 12646, sec 4.8.2).
NEC SpectraView Profiler software asks the user to define the parameters white point,
light intensity, gamma, etc to perform the calibration according to the defined parameters.
It also has a specific calibration targets for which the monitor can be calibrated. It contains
a target for soft proofing according to ISO 12646 and ISO 3664 standards where the target
for calibration parameters is in accordance with the specifications of the ISO 12646
standards. Table 2 shows the parameters selected for the calibration and characterization of
the monitor display.
Table 2 Preset target settings in NEC SpectraView Profiler software
Display Type LCD

Calibration method Hardware calibration (monitor LUTs)
Calibration settings ISO3664 and ISO 12646
Profile settings LUT based (accurate)
CIE Daylight standard D50
Tonal response Curve L* (recommended)
Specify White and black luminance
White 160 cd/m2
Black Minimum neutral
Profile type 16 bit LUT based
Chromatic adaptation CAT02
After selecting the target the software asks for a monitor profile name and consequently
the measurement process were started. According to the defined parameters the software
(SpectraView Profiler) projects colour patches with certain RGB values that are measured
using the X-rite EyeOne Pro spectrophotometer. Depending on the measurement data
appropriate adjustments are made automatically until the difference between the measured
parameters and the target parameters is achieved minimum. Consequently, an ICC monitor
profile is created at the last stage of the calibration process defining device independent
191
Paper C
values (CIEXYZ) that correspond to a given set of device dependent numbers and vice
versa.
After calibrating and creating the monitor profile, the profile was validated in order to
check the objective quality of the generated profile in terms of calculating ΔE*ab colour
difference between given reference values and the corresponding measured values. After
successfully validating, the ΔE*ab colour difference is shown in form of a bar graph and the
profile is automatically applied as the system profile.
Subsequently, the viewing booth settings were checked. ‘Check viewing booth’ option was
selected in the review column ‘Viewing booth and monitor comparison’ and measurements
were performed.
Figure 1 Schematic diagram of the position of the measuring instrument for viewing booth
measurement
To record the measurements, the measuring instrument was held parallel to the viewing
booth facing towards the plane of viewing as shown in Figure 1. Instructions on the screen
were followed. Viewing booth conditions were checked against the P2 viewing condition
mentioned in ISO 3664. In order to reduce the difference between the measured values and
the target values ‘Adjust viewing booth’ tab was selected to make the necessary
adjustments. In the ‘Adjust viewing booth’ tab, for ISO reference values, ‘viewing booth
and monitor comparison’ was selected. It performs the necessary adjustments in the
viewing booth (for light intensity, etc) to match the ISO 3664 P2 viewing condition
requirements. The viewing booth measurements were performed using X-rite EyeOne Pro
with the ambient light adaptor.
192
Paper C
Ambient light condition setup and adjustment

The room used for the soft proof setup has a neutral and matt colour according to ISO 3664
(sec 4.3.4). To setup the ambient lighting conditions according to ISO 12646 standards for
soft proofing, daylight tubes named ‘OSRAM L58W/950, LUMLUX de LUXE, Daylight’
were setup in the room. These tubes simulate Illuminant D50 and the light intensity can be
varied as per the requirements. Therefore, in order to vary the intensity of these tubes a
varying knob facility was introduced in the room with which the ambient light intensity in
the room could be controlled.
SpectraView Profiler software was used to perform the ambient light measurements.
‘Ambient light’ option was selected in the review column and measurements were
performed as per the instructions given by the software. To perform the measurements, X-
rite EyeOne Pro spectrophotometer with the ambient light adaptor was used for measuring
the ambient light in the room. The measurement instrument was held straight parallel to the
display facing towards the room, with the monitor switched off. Furthermore, Figure 2
illustrates the schematic diagram of the position of the spectrophotometer to perform the
ambient light measurements.
Figure 2 Schematic diagram of the position of the measuring instrument for ambient light
measurement
Notice, the geometry of the measurement device is very critical in terms of the measured
values. Small changes of the measurement instrument angle or distance to the light source
can change the measurement results dramatically. The ambient light was measured for
193
Paper C
light intensity and colour temperature. Depending upon the measurements obtained the
light intensity was adjusted using the varying knob facility to be within the tolerance
defined by ISO 12646 standards for soft proofing conditions. After varying the knob to
change the light intensity, measurements were again conducted with a gap of 30 minutes
(time required for the tubes to stabilize after adjusting the light intensity). This procedure
was followed till the time appropriate light intensity was achieved to be within the ISO
12646 standards for soft proofing.
As an alternative a number of software’s like BabelColor CT&A, UDACT, etc can be used
to perform ambient light measurements. BabelColor CT&A uses its Spectral Tool feature
with EyeOne Pro spectrophotometer to measure and evaluate the ambient light conditions.
This software, upon measurements, shows a small graph of u’, v’ co-ordinates which tells
if the measured colour temperature is within the ISO 12646 defined tolerance. BabelColor
CT&A also provide values for Colour rendering index (CRI) and Metamerism index (MI)
of the measured light for P1 condition of measurements. According to ISO 3664: 2009 CRI
and MI are more important parameters to determine the quality of the illumination.
3. Results and discussions
3.1 Monitor display

To analyse and evaluate the display for soft proofing environment the UDACT certifying
software was used. UDACT enables objective, quality oriented and comparable evaluation
for an individual soft proofing display (UGRA 2008).
X-rite EyeOne Pro spectrophotometer was connected to the UDACT software. Instructions
on the screen were followed. UDACT then shows 102 colour patches on the screen of
which measurement values are recorded using the connected spectrophotometer. The
measurements include 21 patches to verify the gray balance. To determine the profile
quality 35 patches are measured. Finally, UDACT analyses the display for the
specifications according to ISO 12646 standards for soft proofing measuring the 46
patches of the Ugra/Fogra Media Wedge. UDACT records the measurements and
generates a report at the end that shows all the details of evaluation. ISO 12646 standards
for soft proofing addresses the factors ‘white point and black point’, ‘gray balance’, colour
gamut’ and ‘uniformity of luminance’ determining the performance of the monitor display.
In the following chapter, the results of the monitor display analysis will be presented.
194
Paper C
White point and black point

According to ISO 12646, white point of the display should be as close as possible to the
calibration target. Table 3 below presents the target and the measured values for white and
black point.
Table 3 White and black point of the NEC display
Target Measured Difference

White point 5000 °K 4991°K 0.5 ΔEab
Luminance 160 cd/m2 159.5 cd/m2
Black point <1.6 cd/m2 0.5 cd/m2
The comparisons between the measured values and the target values in terms of white
point, black point and luminance show acceptable results.
Gray balance
The maximum allowed deviation shall not exceed max ΔC*ab 3 unit and preferably not
max ΔC*ab 2 unit (ISO 12646, sec 4.10). For further details on calculating ΔC*ab and a
general overview of CIE colorimetry see (Hunt 1998)). Table 4 shows the 21 patches for
the gray balance measurement and the performances of the monitor display in terms of the
corresponding measurement and the ΔC*ab deviation. It can be seen that although, the max
ΔC*ab is 2.35 units, it is still within the given tolerance. Flare light could be considered
influencing the measurement on dark colours (e.g. 0%, 5%, 10% and 15%).
Table 4 Gray balance measurements performed and reported by UDACT
% CIELAB (calculated) CIELAB (measured) ΔC*ab

L* a* b* L* a* b*
100 (white) 100 0 0 100.00 0.41 -0.22 0.47
0 (black) 5.01 1.47 0.8 2.72 -0.07 -0.17 1.82
5 9.59 0.85 0.46 7.02 -1.41 -0.20 2.35
10 13.53 0.29 0.14 11.69 -1.39 -0.79 1.92
15 17.63 0 0 16.13 -0.98 -0.93 1.35
20 22.06 0 0 20.96 -0.59 -0.61 0.85
25 26.35 0 0 25.57 -0.40 -0.61 0.73
30 31.09 0 0 30.39 -0.37 -0.52 0.64
35 35.94 0 0 35.43 -0.23 -0.37 0.44
195
Paper C
40 40.84 0 0 40.31 0.06 -0.78 0.78

45 45.41 0 0 45.23 0.04 -0.29 0.29
50 50.37 0 0 50.29 0.15 -0.42 0.45
55 55.37 0 0 55.37 0.06 -0.22 0.23
60 60.38 0 0 60.40 0.28 -0.36 0.45
65 65.03 0 0 64.99 0.52 -0.55 0.76
70 70.05 0 0 70.13 0.13 -0.20 0.24
75 75.10 0 0 75.06 0.39 -0.31 0.50
80 80.15 0 0 80.17 0.54 -0.39 0.67
85 84.82 0 0 84.82 0.37 -0.49 0.62
90 89.87 0 0 89.81 0.54 -0.53 0.76
95 94.93 0 0 94.95 0.23 -0.33 0.40
Average 0.80
Max 2.35
Colour gamut
The colour gamut of the display should be such that it totally encloses the colour gamut
produced by the inks specified in the appropriate part of ISO 12647 for which the display
is required to provide a proof (ISO 12646, sec A.2). UDACT checks for the colour gamut
by measuring the Ugra/Fogra Media Wedge 2.0 on the display simulating the
ISOcoated_v2_300_eci.icc profile. The 46 Ugra/Fogra Media Wedge patches are
visualized using the system profile and the corresponding measurements are recorded and
compared with the reference values. The average colour difference between the reference
values and the measurement values is ΔE*ab 0.8 units. The max ΔE*ab 2.8 and related to
black (16 0 0) which is not unexpected due to the results of the gray balance task which
show the largest colour differences in dark colours. The entire table including the reference
values, measurement values and the calculated colour difference are provided in the
appendix. For illustration purposes Figure 3 shows the top down projection of the colour
gamut of the monitor profile (wireframe) and the ISOcoated_v2_300_eci.icc profile (solid)
in the CIELAB colour space. In this presentation it can be seen that the boundary of the
ISOcoated_v2_300_eci.icc colour gamut is clearly within the gamut of the monitor profile.
196
Paper C
Figure 3 Top down projection of the colour gamut of the monitor profile (wireframe) and
the ISOcoated_v2_300_eci.icc profile (solid) in the CIELAB colour space
On the other hand Figure 4 illustrates the horizontal projection view of the four CIELAB
planes a+, a-, b+ and b- (clockwise) of colour gamut of the monitor profile (wireframe)
and ISOcoated_v2_300_eci.icc profile (solid). It can be clearly observed that the monitor
colour gamut is big enough to simulate the ISO coated printer profile.
Figure 4 Horizontal projection view of the four CIELAB planes a+, a-, b+ and b-
(clockwise) of colour gamut of the monitor profile (wireframe) and
ISOcoated_v2_300_eci.icc profile (solid)
197
Paper C
Uniformity of luminance
Finally the results of the uniformity of the luminance are presented. As previously
mentioned the uniformity will be determined displaying white, grey and black images each
filled the entire screen. Minimum 9 points of the image area of the screen shall be
measured, for each level (Figure 5). According to the ISO 12646 standard requirement the
white image consist of the maximum value in each channel Red, Green and Blue (255 for 8
bit). Then the grey image should have about half of the maximum value in each channel
(127 for 8 bit), and finally the black should consist of approximately a quarter of the
maximum value in each channel (which is e.g. 63 for 8 bit) but shall be greater than 10 %
of the maximum digital code value (which is 26 for 8 bit). Table 5 shows the uniformity of
luminance measurements performed on the display.
Figure 5 Measurement positions on the monitor display
Table 5 Luminance measurements performed on the display at 9 different positions
Luminance measured in cd/m2

Position 1 Position 2 Position 3
White 158.21 156.94 154.07
Grey 30.14 30.22 31
Dark Grey 7.28 7.34 7.66
White 161.75 159.29 155.23
Grey 30.72 30.88 30.77
Dark Grey 7.52 7.5 7.6
198
Paper C

White 173.18 166.46 166.97
Grey 32.8 32.15 32.91
Dark Grey 7.97 7.84 8.17
Figure 6, Figure 7 and Figure 8 show the result of the uniformity check for white, grey
and dark grey neutrals according to ISO 12646 standards for soft proofing. According to
ISO 12646, sec 4.4, the luminance of the display (measured at 9 different locations figure
10) shall be within 10% of the luminance of the measurement made at the centre of the
display. Position 5 (on X-axis) in the graphs below is the centre of the display.
Figure 6 Uniformity of the display for white neutral
Figure 7 Uniformity of the display for grey neutral
199
Paper C
Figure 8 Uniformity of display for dark grey neutral
From the graphs it can be observed that the luminance values varies at 9 different positions
on the display. It can be noticed that the luminance at position 7, 8 and 9 (lower part of the
monitor display) gives the highest values. However, all the measurements are within the
10% tolerance defined by ISO 12646 standards.
3.2 Viewing booth and ambient lighting conditions

To analyse and evaluate the viewing booth and ambient lighting conditions, measurements
were performed using the X-rite EyeOne Pro spectrophotometer with the ambient light
adaptor. The measured value was then compared against the P2 viewing conditions defined
in ISO 3664:2009 standards. Table 6 and Table 7 show the target and measured viewing
booth and ambient light conditions.
Table 6 Viewing booth target and the measured values
Illuminant
Viewing Illuminance
Colour Chromaticity
Booth lux
Temperature Tolerance (u’v’)
Target 5000 °K ≤ 0.005 500 ± 125
Measured 4995 °K 0.00203 480
200
Paper C
Table 7 Ambient light condition target and measured values
Ambient Illuminant
light Colour Illuminance
condition Temperature
¼ monitor white
Target 5000 °K
point luminance
Measured 4853 °K 103 lux
3.3 Soft proofing image data

Implementation of the ISO standard 12646 alone does not guarantee that a displayed (soft
copy) image will match the colour of the same image produced on the hard copy in the
viewing booth (ISO 12646). To obtain a colour match, colour transformation is required to
convert the colour data format from the printer colour space to the display colour space.
This transformation is mainly done by means of a colour management system using the
appropriate ICC profiles. ICC profiles define the relationship between the device specific
colour space and the profile connection space (PCS) (ISO 12646:Annex A.1). Therefore, to
obtain a colour match between hard copy and soft copy proofs using ICC colour
management, the soft proof is obtained by applying the image data, the combination of the
device to PCS transforms of the print output profile and the PCS to device transforms of
the display profile.
In this study, Adobe Photoshop CS4 software was used to perform the colour
transformation. Visual Print reference (VPR) images were applied to simulate the
ISOcoated_v2_300_eci.icc profile. These images were compared to evaluate the visual
match between the VPR hard copy in the viewing booth and the soft copy on the display.
To apply necessary simulation profiles a VPR image was opened in Photoshop and the
menu ‘View > Proof setup > Custom’ selected. In the ‘Customize Proof Condition’
window the appropriate simulating profile was selected. In this case, simulating profile
was the CMYK output profile that needed to be soft proofed (ISOcoated_v2_300_eci.icc
profile). Image background was selected as neutral gray (similar to the viewing booth
background). ‘Simulate paper white’ option was tick marked to simulate the paper white
and the images were then compared with the hard copy in the viewing booth.
201
Paper C
For the soft proofing task Adobe Acrobat Pro software can also be used to display the soft
copy image on the monitor display to compare with the corresponding hard copy image in
the viewing booth.
4. Conclusion
In this paper methods and procedures to setup a soft proofing station are discussed. ISO
12646 standards for soft proofing were successfully implemented. The NEC
LCD2180WG-LED monitor display used in this paper has a very wide colour gamut and is
performing visually uniform. The appropriate calibration and characterization procedure
has been applied. Gamut plots in Figure 3 and Figure 4 showed that the monitor display is
able to simulate printer profile colours. This was verified further by measuring the
UGRA/FOGRA step wedge simulated with ISO coated printer profile on the display using
the EyeOne Pro spectrophotometer. The maximum colour difference obtained was ΔEab =
2.8.
Measurement procedures to measure ambient light using an EyeOne Pro
spectrophotometer with ambient light adaptor are implemented and discussed in this work.
NEC SpectraView Profiler software was used to perform the ambient light measurements.
A number of different software’s can be used to perform ambient light, viewing booth and
display measurements. As an alternative to NEC SpectraView profiler, Babel Color CT&A
software can be used to perform these measurements.
ISO 12646 standards were successfully implemented, as, all the measurements obtained
were within the tolerance level defined. UDACT software was used to evaluate the display
for soft proofing and printer profile simulation according to these soft proofing standards.
ISO 12646 define a wide range of tolerance for ambient light and viewing booth light
intensity measurements. It was observed that, in spite of being within the ISO 12646
standards tolerance level the two images (soft copy image on the display and the
corresponding hard copy image in the viewing booth) might not show an exact visual
match. Therefore, the ambient light intensity and the viewing booth light intensity can be
adjusted (within the tolerances defined by ISO 12646) as per different job requirements to
get the closest possible visual match between the soft copy on the display and the
corresponding hard copy in the viewing booth.
Finally, it is of interest to consider other potential directions for further work in the field
soft proofing. Firstly, to verify the monitor display measurement results of dark colours a
202
Paper C
spectroradiometer could be used. Furthermore, colour measurements of the monitor

display and the viewing booth could be conducted with a spectroradiometer to confirm the
appropriate set up. Moreover, a psychophysical experiment could evaluate the magnitude
of visual differences between the soft proofed image on the monitor display and the hard
copy in the viewing booth in terms of perceptibility and acceptability threshold. Factors
and their magnitude affecting the appearance on the monitor display and the viewing booth
could be investigated.
Nevertheless, using more than one measurement instrument in the soft proofing workflow
the inter-instrument reproducibility has to be considered.
References
CIE15 (2004), 'Colorimetry', 3rd ed (CIE Central Bureau, Vienna).
Gatt, A, Westland, S, and Bala, R (2004), 'Testing the softproofing paradigm', 12th Color
Imaging Conference, p. 187-192.
Gatt, A, et al. (2005), 'Testing the Softproofing Paradigm II', 10th Congress of the
International Colour Association, p. 1243-1246.
Hunt, RWG (1998), Measuring colour (3 ed.: Fountain Press).
ISO 3664 (2009), 'Graphic technology and photography — Viewing conditions', third ed.:
Geneva: ISO [www. iso. org].
ISO12646 (2008), 'Graphic technology – Displays for colour proofing – Characteristics

and viewing conditions', Geneva: ISO [www. iso. org].
JUST (2008), 'Normlicht, Color communicator, Help manual'.
NEC (2005) LCD2180WG-LED technical background and feature overview', Display

Solution <http://www.necdisplay.com/ >, accessed 2. July 2010.
NEC (2008), 'SpectraView Profiler Version 4.1, User’s manual'.
203
Paper C
Nussbaum, P, Sole, A., and Hardeberg, Jon Y. (2009), 'Consequences of using a number of
different color measurement instruments in a color managed printing workflow', TAGA
Proceedings.
UGRA (2008), 'UDACT – Ugra display analysis and certification tool, Help manual'.
UGRA (2009), 'PSO certification', <http://www.ugra.ch/pso-certification.phtml>, accessed

6. July 2010.
Appendix
Reference Measured ΔEab

L a b L a B
1 55 -37 -50 55.4 -36.1 -50.2 1
2 66.9 -24.7 -37.1 66.8 -24.3 -37.2 0.5
3 79.7 -12.5 -21.8 79.8 -12.4 -21.7 0,1
4 48 74 -3.0 48 73.6 -3.7 0.8
5 60.8 50.6 -6.7 60.7 49.8 -6.8 0.8
6 76.4 25.8 -6.9 76.4 25.5 -6.7 0.3
7 89 -5 93 88.5 -4 92.5 1.2
8 90.3 -4.7 62.6 90 -3.9 62 1.1
9 92.2 -3.5 31.1 92 -3.2 31.3 0.4
10 53.1 37.7 28.9 53.2 36.7 29 1
11 41.5 22.7 16.8 41.6 22.3 17.1 0.5
12 31.9 40 24 31.8 40.6 24.4 0.8
13 32.5 44.5 -1.8 32.3 44.7 -1.9 0.4
14 51.3 1.3 44.5 51 0.9 44.6 0.5
15 34.6 -36.4 13.9 33.9 -38 13.5 1.7
16 36 -26.2 -20.9 35.3 -27.1 -22.6 2.1
17 20.9 9.6 -23.6 19.9 9.1 -25.2 1.9
18 89 0 -1.9 89 -0.3 -2 0.3
19 82.8 0 -1.7 82.8 0 -2.1 0.5
20 69.3 0 -1.4 69.2 -0.6 -1.3 0.6
21 54.1 0 -1.0 54.2 -0.6 -1.3 0.7
22 36.6 0 -0.5 36.2 -0.6 -1 0.9
23 16 0 0 14.2 -1.4 -1.5 2.8
24 24 22 -46 23.1 21.7 -47.8 2
25 40.9 17.9 -36.6 40.6 17.8 -37.6 1.1
26 63.7 10.3 -23.8 63.8 10.2 -23.9 0.2
27 47 68 48 47 67.7 48 0.3
28 58.5 47.1 37.9 58.5 46.4 37.9 0.7
29 74.2 22.9 21.4 74.1 22.3 21.2 0.6
30 50 -65 27 49.9 -66.2 26.6 1.3
31 62.1 -39.8 21 62.1 -39.5 20.9 0.4
204
Paper C
32 77 -19.1 11 77.1 -18.9

10.9 0.2
33 71.2 18.9 17.2 71 19.1
17.2 0.4
34 71.2 22.1 73.1 70.8 22.4
72.5 0.8
35 47.7 71.2 16.2 47.7 7116.3 0.2
36 38 55.4 -20.9 38.1 55.6
-21.1 0.3
37 73.7 -22.8 67.6 73.2 -22.3
67.1 0.8
38 52.3 -52.3 -20.1 52.5 -53.3
-20.7 1.2
39 43.3 -17 -48.6 43.1 -17.3
-49.9 1.3
40 95 0 -2 94.9 -0.3
-1.9 0.3
41 88.5 -0.4 -3.1 88.6 -0.6
-3.4 0.3
42 82 -0.9 -4.1 81.9 -1-4.1 0.1
43 67.7 -2 -4.4 67.8 -2.3
-4.2 0.4
44 52.2 -2.5 -3.5 52.3 -2.4
-3.9 0.4
45 37.5 -3.9 -3.1 37.1 -4.3
-3.6 0.8
46 26.3 -6.8 -3.4 25.4 -8.2
-4.3 1.9
Avg 0.8
Measurements performed and recorded by UDACT for colour gamut
205
Paper C
206
Paper D
Paper D
Peter Nussbaum and Jon Yngve Hardeberg

in Coldset Offset Newspaper Print
Color Research & Application,

Article first published online: March 8th 2011, DOI: 10.1002/col.20674
Wiley.
207
Paper D
208
Paper D

in Coldset Offset Newspaper Print
Peter Nussbaum and Jon Y. Hardeberg

P.O.Box 191
N-2802 Gjøvik
Norway
peter.nussbaum@hig.no
jon.hardeberg@hig.no
Keywords: Process control, print quality assessment, newsprint, psychophysical methods,

colour measurement, colour management
Abstract
This paper aims to investigate print quality in newspaper print by considering the
appropriate calibration standard and applying colour management. In particular, this paper
examines the colorimetric properties of eight Norwegian newspaper printing presses, in
order to evaluate the relevant colour separation approach, either by applying custom
separation profiles or by using an industry standard profile. The key method underlying the
work described here relies on obtaining colour measurements to determine the repeatability
of each participant in terms of colour differences. Furthermore, the variation between the
eight newspaper printing presses and the variation according to the colorimetric values of
the ISO 12647-3 standard are important parts of the quantitative evaluation. Based on the
colour measurements two custom ICC profiles were generated and an industry standard
profile “ISOnewspaper26v4.icc” was also used. The first custom profile was generated
using averaged colour measurement data set from a test print run, and the second using a
data set averaged between measured data and the characterization data set “IFRA26.txt”
provided by IFRA. These three profiles were applied to four test images, which were then
printed by the eight newspaper printing presses. A psychophysical experiment was carried
out to determine the ‘pleasantness’ of the reproductions, which were produced using the
three profiles. The results of the study show the performance of the appropriate profile,
209
Paper D
which is applied to the eight newspaper printing presses to obtain significant best print
quality. Eventually the results demonstrate the fact that the print variations in colours
between the eight printing presses are larger than the difference between the custom and
the standard profiles. Hence, the print variations and not the profile selection may have
determined the visual print quality. Therefore the study reveals the importance of adopting
to preserve equal results among different newspaper printing presses.
1 Introduction
Although process control parameters for the production of half-tone colour separations,
proofs and production prints, are clearly defined in ISO 12647-31, often, newspaper
printing processes show significant variations which affect the appearance of print.
Essentially, there are two different approaches considering printing press convention,
namely optimized or standardized press behaviour. A fully optimized press aims at
maximizing its capability in terms of lowest possible dot gain, highest ink densities and
best contrast that the individual printing press can achieve, without considerations of any
external specifications or standards. Such individual parameters can create unique press
conditions, which requires custom ICC profiles for creating the appropriate separations. On
the other hand, an optimized press condition can be considered as a press consistency
where certain predefined parameters (according to standards) over time (repeatability) and
space (uniformity) are preserved. Another approach is to make the press conform to a
certain reference or standard such as ISO 12647-3. By using the second approach and by
standardising the behaviour of the press, industry standard ICC profiles can be used. In
terms of aiming for a common print appearance across printing plants (e.g. preserving print
appearance of an ad campaign) the second approach including consistency will be the most
suitable one to ensure a predictable and equivalent print result.
In 1995, the Norwegian Newspaper Publishers’ Association (NAL) founded its
subsidiary NADA AS. NADA’s primary goal has been to establish a standard for digital
advertisement delivery in Norway, in order to simplify the process of creating and
transmitting digital advertisements from producer to newspaper. Furthermore, NADA has
also been responsible for the generation of three custom newspaper ICC profiles in the
period of 2000 to 2004. NADA highly recommends application of these profiles in the
national newspapers printing process. However, these custom profiles have two common
210
Paper D
characteristics in terms of their parameters. Firstly, the number of colour measurements for
generating the profiles is very small, and secondly, the degree of GCR (Grey Component
Replacement) is rather low considering newspaper print. Therefore, the performance of the
profiles and the corresponding print quality has been considered as not satisfactory and
demonstrates the need for further revision.
Recently, NADA and eight of the largest Norwegian newspaper printing plants started
a project to evaluate their common print quality and print control.
The aim of the presented work is to evaluate eight newspaper printing plants in terms
of their conformance to specified values in accordance with the requirements of ISO
12647-3. Furthermore, the assessment of each individual printing press and the variation
within the 8 participants are important parts of this study to evaluate the appropriate colour
separation approach, either by applying custom separation profiles or by using industry
standard profiles such as ISOnewspaper26v4.icc.2 In order to obtain the defined goals,
NADA in collaboration with the Norwegian Color Research Laboratory has carried out
this print quality project. The following newspaper printing plants participated in the
project: Dagblad-Trykk AS, Schibsted Trykk AS, Nr 1 Trykk AS, Fædrelandsvennens
Trykkeri AS, Adressa Trykk AS, Bergens Tidende AS, Aftenbladet Trykk AS, and Halden
Arbeiderblad AS.
After the present Section introducing the work, Section 2 gives an overview of the
experimental method including information about the quantitative evaluation and the
psychophysical experiment. In Section 3 we present and discuss our results, before
concluding in Section 4.
2 Experimental method
has been concluded as a very complex issue.3-6
Principally there are two different approaches to assessing print quality. The first
approach is by measurement, using instruments to determine values for the various quality
factors. The second method is based on observation, using psychophysical experiments to
gather the judgement of human observers. For instance, the rank order method is a robust
approach where observers are asked to rank the image samples in order, from best to worst,
along an attribute defined by the instructions, such as pleasantness. This method is based
on Thurstone’s «law of comparative judgement».7,8
211
Paper D
Considering the print quality evaluation in this study both approaches have been applied,
using quantitative analyses based on colour measurements, and using psychophysical
experiments (Figure 1).
Figure 1: Overview of method.
2.1 Quantitative evaluation
The purpose of the test procedure described here is to determine the short and long term
repeatability of each participant of the project. Furthermore the variation between the eight
newspaper printing processes and the variation according to the colorimetric values of the
ISO 12647-3 are important parts of the quantitative evaluation. To compare the current
printing conditions the test target ”ECI2002R CMYK.tif” and the “IT8.7-3 CMYK
Target.tif” have been printed and measured. The test target “ECI2002R CMYK.tif” is
corresponding to the current characterization data set, which were used for generating the
Norwegian newspaper ICC profile. Although a number of parameters (e. g. ink set colours,
dot gain etc.) are required to define the appropriate printing conditions the only data
available for the colorimetric comparison between the present press conditions and
characterization data set were the colorimetric data as a part of the private tag in the
present Norwegian newspaper ICC profile (NADA_avis_vjanu2004).
On the other hand the test target “IT8.7-3 CMYK Target.tif” corresponds to the
characterization data set “IFRA26.txt”.9 The ”ISOnewspaper26v4.icc” is the IFRA
newspaper printing standard profile, which is based on the characterization table
"IFRA26.txt” and is valid for the following conditions in accordance with the international
standard ISO 12647-3.
212
Paper D
Furthermore, the colour measurement data measured on the IT8.7-3 CMYK Target from
the first Test Print were the starting point for generating a number of ICC profiles with
different separation settings.
Considering the test print, each printing plant has been asked to use the Norwegian
newspaper production method to perform the printing task.10 In particular, the method
requires 0.90 density for the solid primary colours cyan, magenta and yellow and 1.10
density for black. Furthermore, the setup of the Raster Image Processor (RIP) software has
to be done according to ”linearized newspaper production” agreement. For more details on
Norwegian linearized newspaper production method see the guidelines of NADA10.
In general, a RIP converts information of a page layout such as digital structure in images,
text, graphic elements, and positioning commands into a bitmap file. Furthermore, the
significant process here is the screening and the generation of the appropriate data for
addressing the output equipment, for example the computer to plate (CTP) device.11 All
participants used equal type of newspaper substrate of 45g/m2 grammage.
For the evaluation, the colour measurement conditions were according to ISO 1365512,
the measurement geometry used was 45/0 with a 2º observer angle and CIELAB system.
For documentation, all colorimetric measurements were made on white backing using
Spectrolino/Spectroscan. Considering colour difference calculations, ΔE*ab values were
computed between individual measurement data set and the characterization data set
"IFRA26.txt”. As the colorimetric variation tolerances in ISO 12647-3 are defined in
CIELAB ΔE*ab, only this colour difference metric has been applied in this project. The
arithmetic mean, 95th percentile, and maximum of the resulting colour difference
distributions were then computed.
Although ISO 12647-3 is focusing on colorimetry, the current Norwegian newspaper
production method is still exclusively density control oriented. Therefore the Norwegian
Newspaper Publishers’ Association’s required density measurements and density
evaluation in this project. The density values mentioned above were adopted from the
former ISO 12647-3:1998 standard. However, the density measurement evaluation in this
study were explicitly used for within sheet uniformity control according to the
Association’s requirements and only used for informative purposes. The IFRA Special
Report 2.37 specifies the density tolerances with ±0.10 density.13 The density
measurements using SpectroEye were performed on black backing with DIN E and
polarisation filter. The density deviation tolerance for densitometer measurements is ±0.01
according to DIN 16536-2.14 We measured the 50% patch 10 times and the variation was
213
Paper D
less than 0.01 densities for all process colours. Therefore the repeatability analysis of the
SpectoEye can be considered as satisfying.
2.2 Psychophysical Experiment
The aim of the psychophysical experiment was to determine how pleasant the reproduction
of a newspaper print was considered to be, when compared to the remaining newspaper
reproduction prints. The observers were then asked to rank them in ascending order from
best to worst.7,8 Note that the two test print runs were carried out in this project with six
months of time interval. According to the colour measurements of the first test print, a
number of ICC profiles with different separation settings were generated. An expert panel
determined the two appropriate colour separations, which were then used to carry out the
psychophysical experiment using the test prints of the second test run.
A total of two custom profiles and one industry standard ICC profile were applied to
each test image (by using the relative colorimetric rendering intent) and printed in all the
eight newspaper printing plants.15 The two custom profiles characterize the specific
Norwegian printing conditions whereas the standard ICC profile considers the conditions
in accordance with the international standard ISO 12647-3. The three applied ICC profiles
are:
• NADA X: Custom profile: Is the ICC profile based on the characterization data set
“Average80”. The characterization data set “Average80” is created using
the measurement data, which is the average data of the measurement values
of all eight newspaper printing plants which results in 80 measurements (8
printing plants x10 measurement each) of the IT8.7-3 CMYK testchart
printed. The data have been averaged according to the weighted method,
which ensures that less significance is attached to values with particularly
large deviations when averaging.
Profiling tool: GretagMacbeth ProfileMaker 5.0.8,
Separation setting: Acromatic: MaxK, TIL: 240%, starting point: 5
• NADA Y: Custom profile: Is the ICC profile based on two characterization data sets,
which have been averaged according to the weighted method, which
ensures that less significance is attached to values with particularly large
deviations when averaging.
The first characterization data set (“Average80”) is based on the average
data of the measurement values of all eight newspaper printing plants,
214
Paper D
which results in 80 measurements (8x10 measurements) of the IT8.7-3

CMYK test chart printed. The second characterization data set is based on
the characterization table “IFRA26.txt”9 valid for the reference printing
conditions relating to the international standard ISO 12647-3.
Profiling tool: GretagMacbeth, ProfileMaker 5.0.8,
Separation setting: Acromatic: MaxK, TIL: 240%, starting point: 5
• NADA Z: Industry standard profile, ISOnewspaper26v42: Based on the
characterization table “IFRA26.txt” valid for the following reference
printing conditions relating to the international standard ISO 12647-3.
The images were simultaneously viewed side by side in a viewing cabinet, as shown in
Figure 2 and no anchor stimuli were used as a reference. The viewing set up was based on
the standard condition of the graphic art industry ISO 3664, vertical geometry 45°,
background grey, GretagMacbeth Judge II lighting cabinet with light source D50 and light
intensity 100% (1035 lux).16 The images were viewed in a darkened room (lights off) and
the images, approximately 10cm x 15cm in size, were viewed from a distance of
approximately 65 cm.
Figure 2. Viewing arrangement for psychophysical experiment.
Four test images were chosen to contain a range of different types of pictorial content,
tonal and chromatic variety as shown in Figure 3. The chosen images represent typical
image types used in daily coldset offset newspapers. sRGB is the original colour space for
all four images. Before the images were converted to CMYK using the different ICC
profiles the provided images were verified on a calibrated and characterized monitor
including soft proofing with the three printer profiles applied.
215
Paper D
Figure 3. Four test images: camera, car, portrait and flag. The images “camera” and
“portrait” are reproduced with the permission of Ole Jakob Bøe Skattum. The images
“car”and “flag” are reproduced with the permission of Verdens Gang (VG).
A total of 25 observers (14 experts and 11 naïve) with normal colour vision and an age
range of 20 to 58, took part in the experiment. Each observer was required to assess four
different images. Each image is reproduced by three different separation algorithms and
printed in eight different printing plants. Note that the assessments were carried within
each individual coldset offset newspaper press. The law of comparative judgement was
applied and the method used was rank order.17
The observer’s task was to rank the three prints in the viewing cabinet in order, from
best to worst, in terms of preferred pleasantness. In a study by Morovic18, pleasantness is
defined as the reproduction’s correspondence with preconceived ideas of how a given
image should look according to an individual, in terms of contrast, colour, sharpness, etc.
3 Results and discussion
3.1 Results of the quantitative evaluation
We received 20 Test Prints (printed on one side only) from each of the eight participants.
Subsequently we measured a total of ten test charts (IT8.7-3 CMYK Target) from each
participant to evaluate their colorimetric properties. The test charts have been chosen
according to the density control on solid CMYK bars. 5 density measurements have been
performed on solids across the paper width (Figure 4).
Subsequently the 5 densities across the paper width from all 20-test prints in each colour
have been averaged. Table 1 shows the density values from each participant on Test Print 1
(column 1) and Test Print 2 (column 2). It can be seen that in Test Print 1 all participants
showed an excellent performance within the IFRA density tolerance ±0.10 (except for
“Aftenbladet Trykk” in yellow 0.77 density). However in Test Print 2, except for
“Dagblad-Trykk”, “Aftenbladet Trykk” and “Halden Arbeiderblad” the density
216
Paper D
performance was rather poor and the density values for some of the colours are clearly
outside the IFRA density tolerance specifications. Notice, a high density value in black for
“Fædrelandsvennens Trykkeri” has been observed.
Figure 4. Test chart IT8.7-3 CMYK including solid bars across the paper width to perform
density control.
Table 1. Density performance from each participant in Test Print 1 (1.TP) and Test Print 2
(2.TP). The yellow marked densities are outside the IFRA tolerance value ±0.10 density.
Participants Cyan Magenta Yellow Black

1.TP 2.TP 1.TP 2.TP 1.TP 2.TP 1.TP 2.TP
Dagblad-Trykk 0,93 0.88 0,98 0.89 0,95 1,0 1,17 1.16
Schibsted Trykk 0,97 0.75 0,92 0.85 0,89 0.7 1,12 0.99
NR 1 Trykk 0,86 0.75 0,97 0.94 0,94 0.75 1,11 1.04
Fædrelandsvennens Trykkeri 0,90 0.92 0,93 0.76 0,88 0.88 1,19 1.49
Adressa Trykk 0,83 0.7 0,89 0.76 0,94 0.71 1,05 1.17
Bergens Tidende 0,96 0.97 0,93 1.01 0,98 1,0 1,10 1.18
Aftenbladet Trykk 0,87 0.89 0,88 0.83 0,77 0.89 1,10 1.07
Halden Arbeiderblad 0,81 0.8 0,81 0.81 0,86 0.86 1,20 1.13
In the next paragraph the results of the variations within the sheet, within the press and
between the presses are reported.
217
Paper D
Considering the solid ink uniformity within the sheet, Figure 5 shows the cyan ink
variations on Test Print 1 and Test Print 2 for all participants. Although the solid ink
uniformity in Test Print 1 (measured 5 times across the paper width) is within the IFRA
density tolerance the variations in test print 2 are much larger and partly outside the IFRA
tolerance. For almost all participants, a correlation can be seen between the large variations
and the average density in test print 2. It is important to point out that the appropriate
density and ink uniformity is an important issue of the printing press calibration and will
determine the print quality. However, we emphasize again that density results are reported
for informative purposes only.
To determine the variation in terms of colour stability over time the quality factor
repeatability is required.19 Figure 6 shows the short-term repeatability performance of each
participant individually according to the measurement results of Test Print 1. The colour
differences were calculated between the mean of the measurements taken (10 press sheets)
and each individual measurement (928 patches per sheet) called as Mean Colour
Difference from the Mean (1).
(1)
Although some of the participants show larger variations, the results of all participants are
still within an acceptable tolerance (Mean ΔE*ab < 1.0 units).
Another property of evaluating the performance of a printing process is by analysing the

long-term repeatability, the difference between Test Print 1 and Test Print 2 (time interval
of six months). The colour differences were calculated between the mean of the
measurements taken from Test Print 1 and the mean of the measurements taken from Test
Print 2. It can be observed that “Dagbladet”, “Fædrelandsvennens Trykkeri” and “Adressa
Trykk” show the largest colour differences in terms of the mean (ΔE*ab 5.09, ΔE*ab 4.92
units and ΔE*ab 4.93 units respectively). This can be explained due to different RIP
settings between Test Print 1 and Test Print 2 as reported by the printing plants. On the
other hand due to the excellent density performance between Test Print 1 and Test Print 2
it is expected that “Halden Arbeiderblad” shows a rather low colour difference (mean
ΔE*ab 2.38 units) between the two test prints.
A further approach to assess the colorimetric properties is to analyse the variations
between the eight newspaper printing processes. Figure 8 presents the colour difference
218
Paper D
between the average measurements of all eight newspaper printing plants and each
individual participant in Test Print 2. Notice, that participant “Fædrelandsvennens
Trykkeri” shows the largest difference (mean ΔE*ab > 4.8 units). Furthermore the max
difference has been calculated for the solid black as expected due to the tremendous high
density value (black density 1.49). The participant “Dagblad-Trykk” has the smallest
difference (mean ΔE*ab 2.1 units).
Figure 5. Solid cyan ink uniformity in Test Print 1 (left chart) and Test Print 2 (right chart)
measured 5 times across the paper width.
Figure 6. Short-term repeatability performance of each participant in Test Print 1.
219
Paper D
Figure 7. Long-term repeatability performance of each participant (six months of time

interval).
Figure 8: Variations between the eight participants according to Test Print 2.
The next paragraph reports the results of Tone Value Increase and the colour measurement
evaluation between the presses and the aim values defined in the ISO 12647-3 standard.
The size of the halftone dots increases during the printing process. This is known as
Tone Value Increase (TVI) or dot gain. It is important to know the TVI characteristics of
the printing process to achieve high print quality. The RIP, if necessary, can adjust a TVI
curve. Ideally, a press calibration starts with aiming the target values on the solid tones in
220
Paper D
the primary colours, e.g. according to the defined CIELAB values given in the ISO 12647-
3 standard. Subsequently, the midtone has to be determined. For example, ISO 12647-3
has defined a tone value increase curve of 26%. However, assuming that the tone value
increase curve in the midtone range indicates, for example, only 22% instead of 26%, the
curve must be raised by 4% in the RIP. Table 2 shows the TVI for the 50% control patch
measured for Test Print 1 and Test Print 2. Note that the measurement results demonstrate
large variation in terms of tone value both within the colours cyan, magenta, yellow and
black, between the printing presses and between the two test prints. As mentioned
previously, each printing plant was required to use the appropriate RIP settings according
to the predefined Norwegian linearized newspaper production method including equal RIP
setting for each primary colour.10 However, in order to obtain a common specified dot gain
(e.g. 26%) in all printing plants the RIP setting has to be adjusted individually for each
printing press and for each colour. It can only be speculated what caused the large
variations.
A further important part of the evaluation is to analyse the variation between all eight
participants and the colorimetric values of the characterization data set "IFRA26.txt”.
According to the results given in Table 3, the participant “Dagblad-Trykk” shows a small
colour difference (mean ΔE*ab 4.12 units) while, the participant “Adressa Trykk” shows
the largest colour difference (mean ΔE*ab 7.67 units). This result agrees with the tone value
increase result for 50% value seen in Table 2 (Test Print 2) which have shown a very small
difference between “Dagblad-Trykk” and ISO 12647-3 and a rather large difference
between “Adressa Trykk” and ISO12647-3 respectively.
Table 2: Tone value increase for the 50% control patch measured on paper substrate with
DIN E with polarisation filter on a black background.
Participants: Tone value % Tone value %

Test print 1 Test print 2
C M Y K C M Y K
Dagblad-Trykk 31 28 22 29 25 25 23 27
Schibsted Trykk 17 18 14 4 23 26 19 10
Nr 1 Trykk 11 22 9 17 16 8 9 10
Fædrelandsvennens Trykkeri 20 15 17 10 20 14 18 26
Adressa Trykk 17 13 14 8 8 6 10 14
Bergens Tidende 18 18 13 15 12 14 14 9
Aftenbladet Trykk 10 8 8 12 11 10 14 13
Halden Arbeiderblad 13 10 15 12 21 16 21 8
NADA_avis_vjanu2004 16 19 16 8 16 19 16 8
221
Paper D
ISO 12647-3:2005 26 26 26 26 26 26 26 26
Table 3. Colorimetric colour differences between each newspaper printing plant and the
characterization data set "IFRA26.txt”.
ΔE*ab Mean 95th Percentile MAX

Dagblad-Trykk 4.12 6.54 10.97
Schibsted Trykk 5.23 7.41 10.66
Nr 1Trykk 6.29 10.00 13.18
Fædrelandsvennens Trykkeri 5.34 9.78 14.44
Adressa Trykk 7.67 12.36 19.57
Bergens Tidende 6.85 11.39 17.00
Aftenbladet Trykk 7.32 12.19 16.03
Halden Arbeiderblad 6.53 10.76 15.66
Looking at the CIELAB values for the primary colours cyan, magenta, yellow and black
specified in ISO 12647-3, the differences between the actual values and the nominal values
must not exceed the tolerances shown in Table 5. Table 4 presents the CIELAB values for
the primary colours from all eight participants, ISO 12647-3 and “NADA average 2007”
(Test Print 2). Moreover the table shows colour differences calculated between the
participants and ISO 12647-3. The green marked values indicate that the colour differences
are within the ISO tolerance of ΔE*ab 5 units. The violet marked values are outside the ISO
tolerance. It is interesting to note that according to Test Print 1 almost all participants show
colour differences within the ISO tolerance. On the other hand it can be noticed that except
for “Dagblad-Trykk” and Halden Arbeiderblad” the colour difference in Test Print 2
exceed noticeable the ISO tolerance for most of the participants. Although it might not be
adequate to relate colorimetric results to density values, a correlation between the large
colour differences presented in Table 4 and large density variations seen in Table 1 can be
observed.
222
Paper D
Table 4: CIELAB coordinates for newspaper according to measurements on white backing

(ISO 12647-3).
Table 5: CIELAB tolerances (ΔE*ab) for the solid tones of primaries according to ISO
12647-3.

Deviation Tolerances 5 5 5 5
Note that all participants have used substrates that meet the tolerances for the colour of the
print substrate (ISO 12647-3). Figure 9 shows the average colour gamut projection of the
Test Print 1 and Test Print 2 and the ICC profile “ISOnewspaper26v4” onto the a*b*-plane
at to the level of lightness L*=50. The data used in this task is based on colorimetric
measurements, which have been analysed and visualised by the icc3d application.20 It can
be seen that the average measurement of the Test Print 1 has the largest colour gamut. On
the other hand the average measurement of all the 8 printing presses in Test Print 2 result
in a smaller colour gamut. The measurement data of the “ISOnewspaper26v4” profile lies
between the two test prints.
223
Paper D
Figure 9. 2D colour gamut comparison of Test Print 1 and Test Print 2 and the ICC profile
ISOnewspaper26v4 onto the a*b*-plane according to the level of lightness L*=50.
3.2 Results of the psychophysical experiment
The following are the results in terms of z-scores, which have been obtained in the
psychophysical experiment. The goal of the psychophysical experiment was to determine
pleasantness of the reproductions made according to different newspaper prints. For each
image and newspaper printing plant, the 3 x 1 matrices of ranking order results for each
observer were arranged over the 25 observers and the raw data from the experiments were
treated statistically to obtain z-scores. The precisions of the experimental results are
described in terms of 95% confidence interval (CI), which is calculated using equation (1)
which consists of the mean (R), standard deviation ( ) and the number of observations
(N). Using case V of the method proposed by Thurstone, the standard deviation of the z-
scores is assumed to be .21
(1)
224
Paper D
Figure 10. Rank order z-scores for each newspaper printing press and separation (The
error bars represent the confidence interval 95% of population distribution).
Figure 10 presents the results in terms of z-scores for each newspaper printing plant.
For the number of images and observations, CI was calculated to be ±0.13. Overall it can
be noticed that “NADA X” performs worst for all participants, except for “Adressa
Trykk”. “NADA Z” performs significantly best for participant “Fædrelandsvennens
Trykkeri”. On the other hand “NADA Y” was ranked significantly best for the participants
“Schibsted Trykk”, “Adressa Tykk”, “Bergens Tidende” and “Aftenbladet Trykk”. Note
that the profiles “NADA Y” and “NADA Z” do not indicate a significant difference either
for “Dagblad-Trykk” nor ”Halden Arbeiderblad” in terms of preferred pleasantness.
Considering Figure 11 the experimental results identify the profile “NADA Y” as the
candidate, which scored best for the images “Portrait”, Camera” and “Flag”. “NADA Z”
performs significantly best for the image “Car”. However, “NADA X” was ranked
significantly worst in terms of preferred pleasantness.
Table 6 gives the ranking of the performance of the three separation profiles for the
four images. As it can be seen the profile “NADA Y” was significantly best except for the
image “Car” and the profile “NADA Z” was second. Figure 12 presents the results in terms
of z-scores for all four images and all eight newspaper participants.
225
Paper D
For the total number of images, newspaper printing plants and observations, CI was
calculated to be ±0.05. Notice, overall, the “NADA Y” profile performed significant best
and the “NADA Z” profile second best.
Figure 11. Ranking order z-scores for each type of image and separation (The error bars
represent the confidence interval 95% of population distribution).
Table 6. Ranking of separation profiles for each image (1=best, 3=worst).

Image Profile “NADA X” Profile “NADA Y” Profile “NADA Z”
Portrait 3 1 2
Camera 3 1 2
Car 3 2 1
Flag 2 1 3
Overall 3 1 2
Figure 12. Overall z-scores for the three separation profiles.
226
Paper D
4 Conclusions and perspectives
In this work, we have investigated eight newspaper printing presses in terms of their colour
variations within the sheet, within the press and between the presses. Although all
participants show an acceptable performance in the short-term repeatability (within the
press), the within sheet variations are rather large especially in the second test print. Except
for one printing press the long-term repeatability is inconsistent and has to be improved.
The investigation between the presses also demonstrates some inconsistency.
Considering the TVI, the results show rather large variations between the primary
colours cyan, magenta, yellow and black, between the printing presses and between the
two test prints. A number of factors can have affected the differences. In the course of our
work, we have found that a common TVI specification in relation to dot gain is missing in
the Norwegian newspaper production method. This can explain the inconsistency in terms
of dot gain between the primary colours and between the printing presses. The RIP settings
among the participants can be questioned. Eventually the print variations observed
between Test Print 1 and Test Print 2 have affected the dot gain too. Another possible
factor, which can contribute to the uncertainty of the printed results is the measurement
technology used in the newspaper production process recommended by NADA. The
Norwegian method proposes to use image-based dot meters not only for measuring and
verifying the screening dots on the printing plates but also for measuring the dot gain on
newspaper substrates.10 A study by Wroldsen et al.22 investigated the measurement
performance of dot meters on newspaper substrate. The results of the repeatability analysis
demonstrated very low confidence using dot meters in newspaper print, compared to using
a colorimetric measurement approach.
Considering the conformance to specified values in accordance with the requirements
of ISO 12647-3, the inconsistency of the density values on the solid primary colours,
especially in Test Print 2, has affected the colorimetric values, CIELAB coordinates.
Hence, for most of the participants the colour difference between the measurements of Test
Print 2 and CIELAB coordinates in ISO 12647-3 has exceeded the ISO tolerance.
Although the quantitative evaluation has demonstrated some obvious shortcomings
there is a large potential for improving the target values of the ISO 12647-3 to obtain a
better coherence between the newspaper printing presses. Nevertheless, to preserve the
daily printing conditions and to match the colorimetric requirements of the adopted
227
Paper D
standard profile it is highly recommended to perform press control according to a well

defined standard e.g. ISO 12647-3.
As seen from the results of the psychophysical experiment the ICC profile “NADA Y”
performs significantly better than the other two profiles, “NADA Z” and “NADA X”.
However, it is observed that the measurement parameters of “NADA Z” do not match the
calibration parameters of most of the participants in Test Print 2. Therefore it might not be
expected that “NADA Z” will perform better then “NADA Y”. Moreover, the colorimetric
difference between the measurement data of “NADA Y” and “NADA Z” results in a
colour difference ΔE*ab 2.5 which will be classified as rather small. On the other hand the
print variations within the eight newspaper printing presses in Test Print 2 are between
ΔE*ab 2.2 and ΔE*ab 5 which means that the print variations in colours between the eight
printing presses are larger than the difference between the custom and the standard profile.
Hence, the print inconsistency and not the profile selection may have determined the visual
print quality. Therefore the outcomes in this study demonstrate the importance of adopting
to preserve equal results among different newspaper printing presses. Standardising
newspaper printing means improved communication with customers, and reducing or
eliminating the costs of dealing with complaints. A result of that will make newspaper
advertising more attractive and therefore improving the market conditions for newspapers
in their competition with other media.
Finally, it is of the interest to consider potential directions for further work in the field of
process control and print quality assessment. Colour difference metrics for image quality
assessment have been widely used for various applications. However, there is not very
often a strong correlation between the objective evaluation and the visual assessment.
Furthermore, the interpretation of the complete image quality assessment considering the
colour difference calculation is dependent on the application and the acceptance.
Eventually the acceptability threshold considering print quality assessment is defined as a
vague concept and one that depends strongly on application and industry; further research
is indeed needed to determine the requirements for acceptable print quality.
Acknowledgments
The author thanks the Norwegian Newspaper Publishers’ Association for their kind
permission to use this material. Furthermore, the author wishes to extend its thanks to the
participating members of the project for their grateful collaboration.
228
Paper D
References
1. ISO 12647-3. Graphic technology – Process control for the production of half-tone
colour separations, proof and production prints – Part 3: Coldset offset lithography
on newsprint. ISO; 2005.
2. IFRA. ISO Profiles Download 2010. Available:
http://www.ifra.com/website/website.nsf/html/CONT_CONS_DL?OpenDocument
&CTDL&E&
3. Hardeberg J, Skarsbø S. Comparing color image quality of four digital presses.
Proceedings of the 11th International Printing and Graphic Arts Conference, 2002;
Bordeaux, France.
4. Nussbaum P, Hardeberg JY. Print quality evaluation and applied colour
management in heat-set web offset. In Proceeedings of the 33rd International
Research Conference of IARIGAI, 2006; Leipzig, Germany. p 399-411.
5. Nussbaum P, Hardeberg JY, Skarsbø SE. Print quality evaluation for governmental
purchase decisions Proceeedings of the 31st International Research Conference of
IARIGAI; 2004; Copenhagen, Danmark. Acta Graphica. p 189-200.
6. Pedersen M, Bonnier N, Hardeberg J, Albregtsen F. Attributes of image quality for
color prints. Journal of Electronic Imaging 2010;19:011016.
7. Handley J. Comparative analysis of Bradley-Terry and Thurstone-Mosteller paired
comparison models for image quality assessment. Proceedings PICS conference,
2001. p 108-112.
8. Gescheider G. Psychophysics, Method, Theory, and Application: Lawrence
Erlbaum Associates; 1985.
9. IFRA26 data set. Characterization data for standardized newspaper Coldset-Offset
printing conditions 2004. Available: www.color.org/IFRA26.xalter
10. NADA. 2010. Support – Lineær avisproduksjon – kort og godt. [Guideline].
Available: http://www.nada.no/support/linprod/index.php,
11. Kipphan H. Handbook of print media: technologies and production methods:
Springer Verlag; 2001.
12. ISO 13655. Graphic technology – Spectral measurements and colorimetric
computation for graphic arts images. ISO; 2000.
13. IFRA. Revision of ISO 12647-3: IFRA; 2005 Report nr Special Report 2.37.
229
Paper D
14. DIN 16536-2. Prüfung von Drucken und Druckfarben der Drucktechnik–
Farbdichtemessung an Drucken. Teil 2: Anforderungen an die Messanordnung von
Farbdichtemessgeräten und ihre Prüfung: Deutsches Institut für Normung; 1995.
15. Sharma A. Understanding Color Management. New York: Thompson Delmar
Learning; 2004.
16. ISO 3664. Graphic technology and photography — Viewing conditions. Geneva:
ISO; 2000.
17. Engeldrum P. Psychometric scaling: a toolkit for imaging systems development:
Imcotek Press, Winchester, Mass.; 2000.
18. Morovic J. Colour gamut mapping. In: Green P, MacDonald L, editors. Colour
Engineering Achiving Device Independent Colours: John Wiley & Sons Ltd; 2002.
p 297-317.
19. Morovic J, Nussbaum P. Factors affecting the appearance of print on opaque and
transparent substrates. Journal of Imaging Science and Technology
2003;47(6):554-564.
20. Farup I, Hardeberg J, Bakke A, Kopperud S, Rindal A. Visualization and
interactive manipulation of color gamuts. Proceedings of the Color Imaging
Conference, 2002; Scottsdale, Arizona. The Society for Imaging Science and
Technology. p 250-255.
21. Thurstone L. A law of comparative judgment. Psychological review
1927;34(4):273-286.
22. Wroldsen M, Nussbaum P, Hardeberg JY. Densitometric and Planimetric
Measurement Techniques for Newspaper Printing TAGA 2008;4.
230
Paper E
Paper E
Maria Wroldsen, Peter Nussbaum and Jon Yngve Hardeberg
A Comparison of Densitometric and Planimetric

Measurement Techniques for Newspaper Printing
Published:
In TAGA Journal of Graphic Technology,
Technical Association of the Graphic Arts,
Volume 4, pp. 101-115,
2008.
231
Paper E
232
Paper E
A Comparison of Densitometric and Planimetric

Measurement Techniques for Newspaper Printing
Maria S. Wroldsen, Peter Nussbaum, Jon Y. Hardeberg

E-mail: Maria.Wroldsen@eidsivaenergi.no, peter.nussbaum@hig.no, jon.hardeberg@hig.no
Keywords: densitometer, dot meter, newsprint, process control
First manuscript submitted to TAGA Journal, May 6, 2007

Revised manuscript submitted to TAGA Journal, November 20, 2007
Abstract
Two types of measurement technologies are used for process control in newspaper
printing, namely densitometric and planimetric technologies. Densitometric measurements
are done with densitometers or spectrophotometers, while planimetric measurements are
typically done with CCD image sensor-based instruments called dot meters. Although
print calibration and process control. In this paper we investigate the statistical relationship
between densitometric and planimetric measurements on newspaper print.
The aim of our project was to investigate whether it was possible to estimate halftone
values measured by a densitometer, from the halftone values measured by different dot
meters. The applied model is based on regression analysis using second order polynomials.
The results are given as estimates of the polynomial parameters, i.e. the polynomials give
the relation between halftone measurements with one of the dot meters and halftone
measurements with the densitometer.
Our statistical analysis showed that due to the large uncertainty of the estimated
parameters, the model does not accurately describe the relationship between the two
measurement technologies. This can be explained in part by the poor repeatability
performance for dot meters applied to newspaper print. Moreover the measurement results
also have shown significant variations within the three dot meters used in this experiment.
Factors affecting the repeatability and determining the performance of the model are
considered and discussed in this work.
233
Paper E
1. Introduction
In newspaper printing essentially two types of measurement technologies are used for
process control, namely densitometric and planimetric measurements. In densitometric
measurements, the optical density is measured, and if needed converted to halftone values,
typically using the Murray-Davies equation. In planimetric measurements, it is attempted
to directly measure the halftone values, that is, the dot area coverage, typically using
devices containing a CCD imaging sensor; such devices are often called dot meters.
Although these two technologies are fundamentally different, they are often used
interchangeably in print calibration and process control, in particular in the Norwegian
newspaper industry (Aasen et al., 2002; NADA, 2007).
This motivated us to investigate whether there is a statistically significant
relationship between halftone measurements on newspaper print done with densitometers
(converted into halftone value with the Murray-Davies-equation) and halftone
measurements done with dot meters, The objective of this study is thus to find out whether
it is possible to convert planimetric halftone measurements into densitometric halftone
measurements and vice versa. Since these technologies are used interchangeably, it is
important to know how to convert from planimetric measurements into densitometric
measurements, to keep the printing process under control and to achieve high print quality.
This study is limited to newspaper printed in coldset offset lithography using AM-
screening. The test-target is printed in three different printing devices with different
process parameters. Factors that possibly could have affected the final print (like printing
parameters) are considered as “noise” and will not be discussed. Our focus is to find
whether there is a statistical relationship between densitometric and planimetric
measurements on newspaper print independent of factors that possibly could have affected
the printing process.
This study also includes repeatability analysis for the measuring devices; three dot-
meters and one densitometer. It is necessary to know the repeatability of the measuring
devices, to indicate the validity of a possibly relationship.
After giving a brief overview of the different measurement technologies in Section 2 and
the state of the art of research discussing their properties and relations in Section 3, we
present our experimental setup and preliminary pre-tests in Section 4. In Section 5 we
234
Paper E
present and dicuss our experimental results, before concluding and proposing ideas for
further research in Section 6.
2. Measurement technologies
The size of the halftone dots increase during the printing process. This is called dot gain. It
is important to know the dot gain characteristics to achieve high print quality. The dot gain
is divided into mechanical and optical dot gain. Mechanical dot gain is the result of growth
during the printing process (Malmqvist et al, 1999). Optical dot gain appears due to
absorption and light scattering in the ink and the paper. This makes the dots seem larger
and darker than they really are. The sum of mechanical dot gain and optical dot gain is
called total dot gain.
Density is the light absorbtion ability of a material. The measurement of density is
done with an instrument called a densitometer and is used to control colours in the printing
process. Density is given by:
Ii
Dink = log10 , (1)
Im
where Im is the reflected light intensity and Ii is the intensity of the incident light (Bergman,
2005). High density corresponds to high absorption.
A reflection densitometer measures the amount of reflected light from a surface. It
consists of a light source to illuminate the sample, optics to focus the light, filters to define
the spectral response of the sample and a detector to monitor the reflected light. The
sample is viewed at 45 degrees from the surface. The reflected light is then converted to
density with a logarithmic amplifier and displayed digitally. The densitometer sees the dot
almost like the human eye and provides an optical density value (Tobias Associates, 2007)
Murray (1936) expressed the relationship between the reflection density of halftone
prints and the dot area R, known as the Murray-Davies equation:
1 − 10 − DR
R = × 100%, (2)
1 − 10 − DH
where DR is the density for a sample and DH is the solid ink density.
Traditionally halftone dot area measurements have been done in laboratories with an
instrument called a planimeter (Romano, 1996). This is the same as an image analyzer. In
planimetric measurements, the dot area coverage is measured by using devices containing
235
Paper E
a CCD imaging sensor. Such devices, designed for measuring printing plates, are often
called dot meters. A dot meter combines a microscope and a CCD imaging sensor.
According to Romano (1996) the major variables in a system like this are image capture,
aperture selection and thresholding. The dot meter analyzes the digital image and decides
what is a part of the dot and what is not based on a threshold. The camera takes a snapshot
of the area being measured and literally counts the number of black and white pixels in the
image. The dot meter is actually measuring the dot area and provides an absolute value of
dot coverage (Colthorpe and Imhoff, 1999).
The focus of the camera is an important factor. The depth of focus is typically less
than 0.2 mm for any such system (Colthorpe and Imhoff, 1999). The dot meter uses the
image histogram and a threshold to calculate dot area. The threshold defines how dark a
pixel should be to be taken into account.
3. Literature review
In the past, several studies regarding dot meter and densitometer measurements have been
done; considering their reliability for different materials (Colthorpe and Imhoff, 1999;
Hsieh et al., 2003), comparisons of the two measurement technologies (Spotts et al., 2005;
Hsieh et al., 2003) and the use of an image analysis system to measure density (Malmqvist
et al., 1993; Brydges et al., 1998). Most of these studies deal with measurement of printing
plates. However, lately densitometers and dot meters are used interchangeably and not only
for printing plates, but also on newspaper print (Aasen et al., 2002; NADA, 2007).
Yule and Nielsen (1951) studied whether halftone values from density measurements
corresponded to real dot areas. They found that halftone value calculated from density
values with the Murray-Davies equation did not correspond to the real dot area coverage,
because the effect of the penetration of light into the paper is usually neglected. Especially
for uncoated papers the density of middle tones increases and multiple internal reflections
from the paper surface increase it still further, so that the usual simple equation relating dot
area to density is not accurate. Their general conclusion is that the relationship between dot
area and halftone density is not nearly as simple as it appears.
Arnaud (2001) compared three methods of image analysis to determine the physical
area of dots on five different substrates, among these uncoated paper and printing plates.
The three devices were all based on an optical microscope. His conclusion was that the dot
gain (mechanical and optical dot gain) is a parameter at least as important as the solid ink
236
Paper E
density in process control. Arnaud states that an image analysis software needs to be
created specifically for the printing industry and that this software should be able to
accurately measure dot area on any substrate (including papers).
Romano (1996) states that measuring halftone dot areas on printing plates with a
video image analyzer is a simple procedure, but tends to be rather subjective. It is very
important to obtain a high quality image. According to Romano, the image quality is
dependent of two criterias; the distance between the histogram’s peaks (contrast) and the
depth of the histogram valley (sharpness) of the 50% tint. Illumination is also an important
factor. When it comes to size of aperture (field of view), Romano states that this is a
critical factor to make accurate measurements. With small aperture size (enclosing only a
few dots), errors can occur when the aperture is randomly placed.
4. Experimental approach
As mentioned previously the aim of this study is to find out whether it is possible to
convert planimetric halftone measurements into densitometric halftone measurements.
Hence, a specific test target consisting of 16 patches (patch size 8x8mm) in different
halftone values for each process colour (CMYK), was designed (Table 1 and Figure 1).
The target was printed in coldset web-offset lithography using AM-screening, in three
different printing plants (three different Norwegian newspapers), namely Bladet
Sunnhordland, NR1 Trykk, and Orkla Trykk. In the following, the test targets are referred
to with the name of the printing plant and a number indicating the sequence number of the
copy. For instance, the test-target used as training-set is referred to as “NR1 Trykk 24000”.
Table 1. Digital halftone values (as specified in the image file) for the test target used in
this study (see Figure 1).
C100% C97% C89% C81% C73% C65% C58% C55%

C50% C45% C40% C32% C24% C16% C8% C3%
M100% M97% M89% M81% M73% M65% M58% M55%
M50% M45% M40% M32% M24% M16% M8% M3%
Y100% Y97% Y89% Y81% Y73% Y65% Y58% Y55%
Y50% Y45% Y40% Y32% Y24% Y16% Y8% Y3%
K100% K97% K89% K81% K73% K65% K58% K55%
K50% K45% K40% K32% K24% K16% K8% K3%
237
Paper E
Figure 1. The test target.
As densitometer, a GretagMacbeth Spectrolino spectrophotometer was used, under the

following setup: Physical filter: Pol, White base: Paper, Illuminant: D65, Observer angle:
2º, Density standard: DIN NB. Three commercially available dot meters were used in this
study (brand names withheld for anonymity). Considering the aperture size and the
treshold method used to segment the digital image into ink and substrate areas the
manufacturer’s default settings have been used in this project.
Given the halftone values measured by one of the dot meters, the aim was to predict
halftone values of the densitometer. The applied prediction model is based on regression
analysis using second order polynomials. The results are given as estimates of the
polynomial parameters, i.e. the polynomials give the relation between halftone
measurements with one of the dot meters and halftone measurements with the
densitometer, as follows:
ydensitometer = axdotmeter
2
+ bxdotmeter + c (3)
In polynomial regression, it is important to avoid over-fitting. Graphs with measurement

data indicated that the relation could be described with second order polynomials; the
scatter plots showed slowly decreasing graphs. Third order polynomials were also
investigated, but the third order terms were extremely small. Hence, the polynomials used
are second order to avoid over-fitting, for details, refer to Wroldsen (2006).
Two limitations of the model were introduced; if the predicted densitometer value
exceeds 100% or is below 0%, the value is clipped to 100% and 0%, respectively.
Empirical correlation coefficients were calculated to indicate whether a statistical
relationship exists between the measurement datasets, as follows:
238
Paper E
r = Correl( X ,Y ) =
∑( x − x )( y − y )
∑( x − x ) ∑( y − y )
2 2 (4)
Because of significant measurement differences between the process colours, it was

necessary to study each of them individually. Furthermore the measurement data were
divided into two sets; a training set to establish the model and a test set to evaluate its
performance. The residuals between the predicted and measured halftone values with the
densitometer (with test set = training-set and test set ≠ training set, respectively) were used
to judge the performance of the model.
Because of significant measurement differences between the dot meters and also
between the process colours for each densitometer-dotmeter-combination, it was necessary
to study both the instrument combinations and the process colours individually (Wroldsen,
2006). The modelling and data analysis were therefore conducted separatly for each dot
meter.
The following method was used in this study to build and test the model (describing
a possible relationship between densitometric and planimetric measurements) for each
combination of instruments: First, three series of measurement data from one test target
were used to establish the model (one model for each process colour). This measurement
data constitutes the training set. Then, the residuals between predicted and measured
halftone values with the densitometer (with the test set being part of the training set and
with the test set totally independent of the training set, respectively) were used to judge the
performance of the model. Some of the test targets in the test set were printed in another
printing plant than the test target of the training set.
4.1 Preliminary repeatability tests
To justify that the densitometer could be used as a reliable representative for all
densitometers, we did a preliminary test with two different densitometers. This was done to verify
whether different densitometers give the same result (in contrast with the dot meters which are
based on thresholds). For this test we used the test-target named “Bladet Sunnhordland 1500”. The
following patches were measured for each process colour: 100%, 81%, 50% and 24% (white base:
paper).
The density values were converted to halftone values using the Murray-Davies-
equation, and the densitometer pretest showed the largest deviations for halftone values
below 24%. This is probably caused by the conversion from logarithmic density values
239
Paper E
into halftone values. Low density values converted to halftone values using the Murray-
Davies equation result in larger variations than high density values. This effect is getting
even more obvious with low solid ink densities, like in newspaper printing. Another
critical factor is the number of decimals used for density measurement. Even though the
densitometer pretest showed some deviations between the two densitometers, only one of
them is used in the analysis. This was necessary to limit the analysis.
The repeatability analysis of the densitometer was satisfying. We measured the 50%
patch 10 times. The variation was less than 0.01 density for all the process colours. The
tolerance density deviation for densitometer measurements is ±0.01 according to DIN
16536-2 (1995).
A repeatability analysis was also conducted for the dot meters. On newspaper print
the 50%-patch for each process colour (CMYK) was measured 10 times with each dot
meter. Based on these measurements, we calculated the average; range (absolute value of
maximum halftone value minus minimum halftone value) and standard devation were for
the three dot meters on newspaper print (note: not printing plates). The repeatability
analysis showed low repeatability for all three dot meters, as shown in Table 2, 3 and 4.
Table 2. Repeatability analysis – dot meter 1.
Dot meter 1 Test-target: “NR1 Trykk 24000”

C50% M50% Y50% K50%
Average 49.70% 46.35% 47.50% 41.05%
Range 3.00% 3.00% 3.50% 1.00%
Standard deviation 0.92% 1.11% 1.00% 0.28%

C50% M50% Y50% K50%
Average 48.86% 48.25% 52.53% 39.69%
Range 5.80% 6.10% 7.00% 2.80%

C50% M50% Y50% K50%
Average 49.15% 42.25% 47.80% 40.40%
Range 4.00% 3.50% 2.50% 2.00%
240
Paper E
DIN 16536-2 (1995) states the tolerance variation of density measurements to be ±0.01.
However, there is no standard dealing with acceptable variations for dot meter
measurements. In accordance with ISO 12647-3 (2005) the optical density for CMY
should be 0.9 and for black 1.1. Outside U.S the 26% tonal value curve is used. This means
that the tone value increase at 40% or 50% should be 26%. Optical solid density 0.9 and
26% tone value curve make ±0.01 correspond to approximately 2% tone value for the
middle tones (see Table 5). None of the three dot meters fulfilled this requirement.
Table 5. Tolerance tone value for the middle tones.
Density Murray-Davies
DH DR Halftone value
0.90 0.90 100.0%
0.48 76.52%
0.47 75.64%
0.46 74.73%
According to the presented results, it is not possible to decide whether this low
repeatability is caused by the measuring devices and/or inhomogeneous halftone values
within one patch. Print irregularities cause noticeable differences in measured halftone
values and reduce the repeatability when the aperture is small. Large screen dots used in
newspapers in combination with small aperture is therefore unfavorable. It is not
unambiguous to decide what is substratum and what is part of a screen dot, especially for
middle halftone values, due to high optical dot gain (see Figure 2).
Figure 2. Image of a 50% halftone value printed on newspaper with magenta ink taken
with a Zeiss Axioplan 2 imaging microscope (© Maria S. Wroldsen).
241
Paper E
As mentioned previously the dot meter analyzes the digital image and according to a
certain defined threshold operation the colour image will be converted into a high-contrast,
black-and-white image. All pixels lighter than the defined threshold are converted to white;
all pixels darker are converted to black. Figure 3 illustrates variations in terms of defining
an appropriate threshold to determine the size of the dot area. Taking a porous and
scattering substrate such as newsprint into consideration the main difficulty is to determine
an ideal threshold.
Figure 3. Treshold variations and the corresponding % dot area. Image A with treshold
100 results in 25% dot area, Image B with treshold 128 results in 42% dot area and Image
C with treshold 150 results in 52% dot area.
5. Experimental Results and Discussion
The calculation of empirical correlation between the densitometer and dot meter values
indicated relationship (see Table 6). Due to the fact that the correlation coefficients were
close to 1, it can be assumed that there must be a correlation between the halftone values
measured by the various instruments.
Table 6. Empirical correlation coefficients
Dot meter 1 Dot meter 2 Dot meter 3
Densitometer, C 0.984 0.986 0.979
Densitometer, M 0.991 0.984 0.982
Densitometer, Y 0.988 0.997 0.991
Densitometer, K 0.986 0.983 0.985
Based on three measurement series of one test target, second order polynomials estimating
the relationship between halftone measurements with the dot meters and the corresponding
halftone measurement with the densitometer were established.
Figures 4, 5, 6 and 7 show measurement data with belonging trendlines for the three
measurement series with dot meter 1, of the test target “NR1 Trykk 24000”; this
constitutes the training set for our model. These graphs show that the trendlines highly fit
the measurement datas. To test this model, we calculated residuals between predicted and
measured halftone value.
242
Paper E
Cya n
100,00
90,00
y = -0,006x 2 + 1,6953x - 4,7409
80,00
R2 = 0,9937
Halftone value in %, (densitometer)
70,00
60,00
50,00
40,00
30,00
20,00
10,00
0,00
0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00
Halftone value in % (dot meter 1)
Figure 4: Relation for cyan (“NR1 Trykk 24000”, dot meter 1)
Ma ge nta
100,00
90,00
2
y = -0,0042x + 1,4406x + 4,4598
80,00
R 2 = 0,9953
70,00
60,00
50,00
40,00
30,00
20,00
10,00
0,00
0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00
Figure 5: Relation for magenta (“NR1 Trykk 24000”, dot meter 1)
243
Paper E
Ye llow
100,00
90,00
y = -0,0049x 2 + 1,468x + 3,606

80,00
R2 = 0,9965
70,00
60,00
50,00
40,00
30,00
20,00
10,00
0,00
0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00
Figure 6: Relation for yellow (“NR1 Trykk 24000”, dot meter 1)
Bla ck
100,00
90,00
2
y = -0,0052x + 1,5301x + 1,5519
80,00
R2 = 0,9967
70,00
60,00
50,00
40,00
30,00
20,00
10,00
0,00
0,00 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00
Figure 7: Relation for black (“NR1 Trykk 24000”, dot meter 1)
The obtained regression polynomials for dot meter 1 are given in Equations 5-8.
Cyan : y densitometer = −0.006x dotmeter1

2
+ 1.6953x dotmeter1 − 4.7409 (5)
Magenta : y densitometer = −0.0042x 2
dotmeter1 + 1.4406x dotmeter1 + 4.4598 (6)
244
Paper E
Yellow : y densitometer = −0.0049x dotmeter1

2
+ 1.468x dotmeter1 + 3.606 (7)
Black : y densitometer = −0.0052x dotmeter1
2
+ 1.5301x dotmeter1 + 1.5519 (8)
The residual between predicted and measured halftone value with the densitometer was
calculated as in the following example with cyan 65%:
• Measured halftone value with dot meter 1: 59.50%

• Measured halftone value with the densitometer: 78.12%
• Predicted halftone value with the densitometer: 74.89%
(relation for cyan; y = -0.006x2+1.6953x-4.7409)
• The residual between measured and predicted halftone value with the densitometer:
3.23%
Table 7 shows the residuals when the test-set is part of the training-set. Because the test set
is part of the training set, it is expected that the differences between the predicted and
measured halftone values are rather small. As it can be seen in Table 7 the variations are
colour and halftone value independent and does not follow a certain trend although cyan
shows the largest variations. However, there is no significant trend for the obtained
variations. Although the model performs well it is important to test the model with another
test-set.
Table 8 shows the residuals when the test-set is not part of the training-set. It can be seen
that the model does not perform that well applying another test-set. Consequently the
differences between the predicted and measured halftone values are larger. Although the
residuals have increased, the variation still does not follow a certain trend.
Table 7. Dot meter 1, densitometer: Residuals in % when the test set is part of the training
set. (Test set: “NR1 Trykk 24000”, 1. measurement series.)
Residuals
Patch Cyan Magenta Yellow Black
100% 0.00 0.00 0.10 0.00
97% 0.13 0.14 0.14 1.54
89% 1.20 1.13 0.85 0.58
81% 3.63 1.61 1.50 0.58
73% 3.23 0.92 0.97 0.04
65% 3.23 2.62 0.87 0.68
58% 0.79 1.33 1.53 0.99
55% 1.60 1.68 3.59 3.73
50% 1.47 0.15 1.49 0.35
45% 0.95 1.86 0.84 1.30
40% 1.19 2.01 1.18 0.37
32% 2.09 0.11 4.46 1.20
245
Paper E
24% 0.78 1.00 0.97 2.66

16% 0.98 1.36 1.51 0.83
8% 0.55 3.77 0.53 0.82
3% 4.25 4.87 1.52 1.09
Table 8. Dot meter 1, densitometer: Residuals in % when the test set is different from the
training set. (Test set: “Orkla Trykk 5000”.)
Residuals
Patch Cyan Magenta Yellow Black
100% 0.00 0.00 0.00 0.00
97% 2.15 1.59 1.66 0.94
89% 2.12 3.24 1.86 3.16
81% 3.25 5.83 2.15 1.46
73% 4.38 5.57 3.78 3.71
65% 5.04 0.40 3.82 4.99
58% 4.82 1.46 4.17 6.04
55% 3.65 0.78 4.07 4.85
50% 1.06 2.66 2.64 1.0
45% 0.16 1.48 4.51 1.75
40% 1.45 1.64 3.99 2.43
32% 2.61 5.86 1.69 1.54
24% 4.55 5.52 3.43 3.07
16% 2.60 2.19 0.59 1.01
8% 0.17 4.09 0.37 0.72
3% 4.91 1.80 4.66 0.48
Residuals between predicted and measured densitometer values were calculated for several
test sets different from the training set. Some of these test-sets where printed in different
printing processes than the training-set.
As stated earlier, 2% can be considered the tolerance variation for dot meters. Some
residuals are larger than 2% and this indicates that the model is not good enough to
describe a relation. The low repeatability is an unfavorable factor.
We have so far presented and the detailed results from the first dot meter only. In the
following we summarize briefly the results for the two other dot meters, for more detailed
information refer to Wroldsen (2006).
For dot meter 2 the obtained regression polynomials are shown in Equations 9-12. The
residuals are larger when test-set is different from the training-set (see Table 10) than for
test-set part of training-set (see Table 9) and indicates that our second order polynomials
246
Paper E
do not satisfactory describe a possibly relationship between halftone measurements with

dot meter 2 and the densitometer. This is partly caused by the low repeatability for dot
meter 2.
Cyan : y densitometer = −0.0061x dotmeter

2
2 + 1.6114 x dotmeter 2 − 2.0895 (9)
Magenta : y densitometer = −0.0051x dotmeter

2
2 + 1.5363 x dotmeter 2 + 1.092 (10)
Yellow : y densitometer = −0.0018 x dotmeter

2
2 + 1.0991x dotmeter 2 + 7.1581 (11)
Black : y densitometer = −0.0063 x dotmeter
2
2 + 1.6296 x dotmeter 2 + 0.3169 (12)
The polynomials for dot meter 3 is are given in Equations 13-16, and the residuals are
given in Table 9 and 10. We see the same trend as for the two other combinations of
instruments. The average residuals are larger when test-set is not part of training-set. As for
the two other combinations, the low repeatability of the dot meter is one factor that makes
our model unsatisfactory.
Cyan : y densitometer = −0.0049 x dotmeter

2
3 + 1.3198 x dotmeter 3 + 13.486 (13)
Magenta : y densitometer = −0.0044 x dotmeter

2
3 + 1.3316 x dotmeter 3 + 10.239 (14)
Yellow : y densitometer = −0.0043 x dotmeter

2
3 + 1.2947 x dotmeter 3 + 11.319 (15)
Black : y densitometer = −0.0047 x dotmeter
2
3 + 1.4000 x dotmeter 3 + 5.5886 (16)
Table 9. Residuals in % between dot meter and densitometer when the test set is part of the
training set. (Test set: “NR1 Trykk 24000”, 1. measurement series.)
Residuals
Dot Max 4.25 4.87 4.46 3.73
Meter 1 Average 1.63 1.54 1.38 1.05
Dot Max 5.88 4.11 3.73 4.80
meter 2 Average 2.27 1.29 1.54 1.32
Dot Max 8.50 4.56 3.74 4.96
meter 3 Average 1.89 2.19 1.88 1.79
247
Paper E
Table 10. Residuals in % between dot meter and densitometer when the test set is different
from the training set. (Test-set: “Orkla Trykk 5000”.)
Residuals
Dot Max 5.04 5.86 4.66 6.04
Meter 1 Average 2.68 2.76 2.71 2.32
Dot Max 4.37 8.39 6.87 6.87
meter 2 Average 2.09 3.65 2.63 2.35
Dot Max 8.94 6.31 10.02 6.17
meter 3 Average 2.93 3.39 4.87 2.44
6. Conclusions and Perspectives
Our statistical analysis showed that due to large uncertainty of the estimated parameters,
the model does not accurately describe the relation between the two measurement
technologies. This can be explained by the poor repeatability performance for dot meters
applied in newspaper print. The repeatability analysis provided, already in the first part of
this study, low confidence using dot meters in newspaper print. None of the three dot
meters fulfilled the requirement of 2% tolerance deviation (note: these are requirements
which are not defined in an official standard). Dot meters are originally developed for
measuring printing plates only.
The residuals between predicted and measured half tone values with the densitometer
increased when the test set was different from the training set, as would be expected.
Moreover, the measurement results have shown significant variations within the three dot
meters. Some factors affecting the repeatability and determining the performance of the
model are listed in this section.
Important factors that impair the use of dot meters on newspaper print:
• Print irregularities cause noticeable differences in measured halftone values and reduce
the repeatability when the aperture is small.
• Small aperture in combination with large halftone dots used in newspapers are
unfavourable.
• Due to high optical dot gain (especially for the middle tones and in newspaper print) it
is ambiguous to decide what is substratum and what is part of a screen dot.
• Large residuals between predicted and measured halftone values for the middle tones
could partly be explained by the high optical dot gain and problems due to
determination of threshold (what is substratum and what is not) in the image analysis.
248
Paper E
Important factors that impair the use of densitometers on newspaper print (when using the
Murray-Davies equation to calculate halftone values):
• The convertion from logarithmic density values into halftone values with the Murray-
Davies equation causes a slowly decreasing graph that makes low density values
converted to halftone values result in larger variations than high density values.
• The effect of this conversion is even more obvious when used with low solid ink
densities like in newspapers.
Based on these results, dot meters are not recommended for halftone measurements on
paper substrates in newspaper printing.
Throughout this project some ideas of further research to investigate a possibly
relation between densitometric and planimetric measurement emerged.
The test target was printed in different printing processes with different solid ink
densities, even though the instructions for printing said K 1.10 and CMY 0.9 in accordance
with ISO 12647-3 (2005). It is difficult to control this in newspaper printing. It would have
been interesting to do the same experiment with a print medium where accurate solid ink
densities are possible. The uncertainty of dot meters and densitometers for use in
newspaper printing is too high. Another type of paper (with lower optical gain) would also
be preferable. To increase the repeatability it is advantageous with coated paper, accurate
solid ink density and finer screen ruling (the aperture size would not be so critical).
Moreover, FM-screening could be used. The reason why we did this experiment with
newspaper in the first place is the increasing use of dot meters in the newspaper industry.
More than one copy of each instrument could have been included in the repeatability
analysis to make any variations between copies become visible.
Image analysis of halftone images would be interesting to investigate the decision of
threshold (what is part of a screen dot and what is not) and to illustrate the percent of
optical dot gain for different halftone values. Different thresholds could be set and the
result (dot area coverage) could be compared to measured values for the different dot
meters. Different size of aperture could also be simulated to observe the influence of
calculated dot area coverage. This experiment could perhaps lead to a recommandation of
optimal size of aperture for different screen rulings; what size of aperture is necessary to
avoid systematical errors?
249
Paper E
Acknowledgments
First, we would like to thank senior lecturer Sven Erik Skarsbø for providing supervision
to this master thesis project (in collaboration with professor Jon Y. Hardeberg and lecturer
Peter Nussbaum). Thanks for your interest, motivation and help throughout this project!
Secondly, we would like to thank associate professor Are Strandlie for valuable help with
the statistical analysis. Finally, we would like to thank Mediebedriftenes Landsforening for
financial support and the newspapers (Bladet Sunnhordland, Romerikes Blad and
Drammens Tidende) for offering to print the test targets.
References
Arnaud, S. (2001). “Measurement of dot area”, In TAGA Proceedings, pp. 685-702
Bergman, L. (2005). “Using Multicoloured Halftone Screens for Offset Print Quality
Monitoring”, Licentiate Thesis No. 1147, Linköping University.
DIN 16536-2 (1995), “Prüfung von Drucken und Druckfarben der Drucktechnik –
Farbdichtemessung an Drucken. Teil 2: Anforderungen an die Messanordnung von
Farbdichtemessgeräten und ihre Prüfung.” Deutsches Institut für Normung.
Colthorpe, S. and Imhoff, G. (1999). “CTP – Why densitometers do not work”, White
paper, Centurfax Ltd and Grip Digital Inc.
Hsieh, Y., Wu, Y. and Lin, W. (2003). “An Expermental Research to Compare Devices for
Measuring Aluminium Lithographic Printing Plates.”, In TAGA Proceedings, pp. 250-273.
ISO 12647-3 (2005). “Graphic technology – Process control for the manufacture of half-
tone colour separations, proofs and production prints. Part 3: Coldset offset lithography on
newsprint.” International Organization for Standardization.
Malmqvist, K., Verikas, A. and Bergman, L. (1999). “Consistency of mechanical dot gain -
a hidden quality parameter”, TAGA 51st Annual Technical Conference, pp 409-417.
Murray, A. (1936). “Monochrome Reproduction in Photoengraving”, J. Franklin Institute,

vol. 221, pp. 721-744.
250
Paper E
NADA (Visited 2007), “Support – Lineær avisproduksjon – kort og godt.” Online:

http://www.nada.no/support/linprod/kortoggodt.html. In Norwegian. (NADA is a company
working with digital ad delivery for Norwegian newspapers, see also
http://www.nada.no/omnada/english.html)
Romano, D. (1999). “The Image Analyzer – A True Dot Area Meter?” In TAGA
Proceedings, pp. 318-334.
Tobias Associates, Inc. (Visited 2007), “Reflection Densitometry,” Online:

http://ourworld.compuserve.com/homepages/tobiasinc/TAPages/bulletin3.htm
Yule, J. A. C. and Nielsen, W. J. (1951). “The penetration of light into paper and its effect
on halftone reproduction”, Taga Proceedings, pp. 65-76.
Wroldsen, M. S (2006). “Densitometriske og planimetriske målinger av rasterstrukturer”,

Master’s thesis, Gjøvik University College, Gjøvik, Norway. In Norwegian.
Aasen, E. Danielsen, Ø., and Bovolden, A. J. (2002), “Halftone measurements in

newspaper print”, Bachelor thesis, Gjøvik University College, Gjøvik, Norway. In
Norwegian.
251
Paper E
252
Paper F
Paper F
Peter Nussbaum, Aditya Sole and Jon Yngve Hardeberg
Analysis of Color Measurement Uncertainty

in a Color Managed Printing Workflow
Accepted for publication

in Journal of Print and Media Technology Research,
The International Association of Research Organizations for the Information,
Media and Graphic Arts Industries,
(iarigai).
253
Paper F
254
Paper F
Analysis of Color Measurement Uncertainty

in a Color Managed Printing Workflow
Peter Nussbaum, Aditya Sole, Jon Y. Hardeberg

E-mail: peter.nussbaum@hig.no, aditya.sole@hig.no, jon.hardeberg@hig.no
Keywords: Color measurement, calibration, color differences, print quality assessment,

color management
Abstract
Since the recent revision of ISO 12647-2 and ISO 12647-7, specifiying the requirements
for systems that are used to produce hard-copy digital proof prints, the use of color
measurement instruments is even more than before required in the printing industry.
Currently, there are many different makes and models of color measurement instruments
used in the industry. Therefore, in a modern color managed printing workflow, most of the
printing houses use more than one color measurement instrument, typically one instrument
in each department (pre-press, press, and post-press).
In this paper, a total of nine commercial spectrophotometers are compared in terms of
measurement uncertainty, precision and accuracy, repeatability and reproducibility. The
BCRA series 2 ceramic gloss tiles are used to confirm the accuracy and repeatability of
these measuring instruments according to the manufacturer’s standards. We focus
especially on inter-instrument and inter-model reproducibility and discuss the effect of
instrument calibration and certification.
For our experimental setup, four different materials are used, one proof print, one
commercial print, and one reference print, along with the BCRA series 2 ceramic gloss
tiles. In a color managed printing workflow the use of more than one instrument can impair
and complicate the color process control due to the color differences between different
measurement devices. The effect of the colorimetric measurement errors due to large inter-
instrument and inter-model variability between instruments used in different parts of the
workflow (e,g, in the printing house, at the customer’s site for inspection, and for
certification) is discussed and demonstrated in this paper.
255
Paper F
1. Introduction
Recently ISO 12647-2 [1] and ISO 12647-7 [2], have defined the colorimetric parameters
for process control in the graphic arts, and tolerances for their acceptance. However, the
color measurement devices used in a production workflow (from the costumer, designer,
prepress, to printing house) may show variations in terms of precision (repeatability,
reproducibility) and accuracy of the measurements made. Furthermore, in the context of
PSO (Process-Standard Offset) certification, test prints and proofs are printed according to
certain aim parameters. The prints and proofs are measured twice, firstly in the printing
house with the instrument of the company, and secondly by the certification body to ensure
that prints are made within the predefined tolerances. In a practical application, if both
measurement devices result in values which qualifies the print or proof as approved, then
nobody will question the reliability (precision and accuracy) of the instruments. Similarly
if both instruments give values that are outside the given ISO production tolerance.
However, if only one of them qualifies the results to pass then you might think of the other
one as a false-positive. Moreover, it lies in the Human’s nature to believe that the
instrument, which does not qualify the print or proof as approved is not performing
appropriate - without taking into consideration that it might be the proof or print which is
in deed produced outside the defined tolerances. However, by improving the instrument
accuracy and reducing the inter-instrument and inter-model agreement will contribute in
reduce inappropriate considerations.
The aim of the presented work is to evaluate the performance of nine color measurement
instruments in terms of precision (repeatability, reproducibility) and accuracy.
Furthermore, the effect of color measurement variability in a color managed printing
workflow will be demonstrated. In particular the result of inter-model agreement
measuring colors on paper substrate will be reviewed.
After this brief introduction, we give some more background information in Section 2, by
illustrating the problem, defining key concepts, and discussing central references. Then, in
Section 3, we describe our experimental setup. In Section 4 we present and discuss our
results, before concluding in Section 5
256
Paper F
2. Background
To illustrate how measurement uncertainty in a printing production workflow may cause

unexpected discussions, Figure 1 shows a simplified diagram of a practical scenario in
which two instruments are used to measure the same target in a color workflow. Given a
certain color patch reference and measuring the patch with the color measurement
instrument of the customer will result in ΔE*ab 3.5. Although the inter-instrument
reproducibility between the customer and print house measurement devices is ΔE*ab 3.0,
the color difference measured with the print house measurement device on the color patch
is almost twice the one measured with the instrument of the customer. Furthermore,
assuming having a certain color difference tolerance of e.g. ΔE*ab 5.0 the result of the print
house measurement of ΔE*ab 6.5 would not be accepted.
Figure 1. Simplified diagram of a practical scenario using two instruments in a workflow

measuring the same target/reference.
According to Berns [3], measurement uncertainty can be divided into two main categories,
precision and accuracy (Figure 2). Precision describes the dispersion of the measurements
taken. On the other hand accuracy describes the distance between the measurements taken
by the color measurement instruments and the actual target value (Figure 3). Accuracy is
affected by systematic errors, which are errors due to different geometry, detector linearity
errors resulting from wavelength.
Precision can be further divided into repeatability and reproducibility. ASTM Standard E
284 [4] defines the repeatability as “the closeness of agreement between the results of
successive measurement of the same test specimen, or test specimens taken at random
from a homogeneous supply, carried out in a single laboratory, by the same method of
measurement, operator, and measurement instrument with a repetition over a specified
257
Paper F
period of time”. On the other hand, changing conditions such as the operator, measuring
instrument, laboratory, or time, gives a measure of reproducibility.
Figure 2. Overview over measurement uncertainty.
Figure 3. Accuracy describes the distance between the measurement and the target and
precision the dispersion of the measurement taken (after Berns' figure on page 95 [3]).
In the past, various studies and research regarding measurement uncertainties and
comparative studies of color measurement instrumentations have been presented. Already
40 years ago Billmeyer [5] presented a work where he studied the comparative
performance of fifteen different color measurement instrument according to their precision
and accuracy. Some years later another research conducted by the same author [6]
concludes that the three instruments tested over en period of seven weeks are essentially
equivalent in the precision and accuracy, measuring a large variety of samples, including
textile. A slightly different approach has Rodgers [7] used in his comparative study of
color measurement instrumentation were he compared not only the inter-instrument
agreement but also the user friendliness of the software and computer interface, vendor
amenability to a long term logistical and maintenance relationship and finally the price. A
quality improvement team considered the most critical parameter to be the inter-instrument
258
Paper F
agreement, followed by the software, the logistical/service relationship, and at last the
price. Another comprehensive set of work was carried out by Wyble addressing the
evaluation of methods for verifying the performance of color measurement instrument such
as integrating sphere and bidirectional devices. Part I is covering the repeatability issue [8]
and part II is addressing inter-instrument reproducibility [9]. However, our main
contribution is the analysis of color measurement variability of using a number of different
color measurement instruments in a color managed printing workflow.
The specifications, methods and procedures to evaluate the performance of color

measuring instruments in terms of repeatability and accuracy are defined in ASTM E2214
[10]. Because, the assessment of the instrument’s performance and the analysis of the color
measurement uncertainty is an important aim of this work, a brief evaluation of the
methods is provided here.
Accuracy describes the conformance of a series of readings to the accepted or true value.
Repeatability defines how well an instrument repeats it’s reading of the same target over a
certain period of time. The assessment of the instrument’s consistency can be tested over
three periods of time. First is the short-term repeatability which is based on measurements
made in succession, second is the medium-term repeatability which can be based on
measurement’s made over a period of hours and finally the long-term repeatability which
is based on measurements made over weeks or longer. The short-term measurements can
be performed either with or without replacement of the measuring instrument from the
color tile/patch to be measured. When measuring without replacement, the tile/patch is left
in place at the instrument’s aperture. This approach might be dependent on the instrument
technology and the user interface. To obtain the most reproducible results, measurements
have been restricted to the central region of the tiles. The color differences are calculated
between the mean of the measurements taken and each individual measurement called as
Mean Color Difference from the Mean [3].
(1)
Reproducibility is a form of repeatability in which one or more of the measurement

parameters have been systematically changed such as the target is different, the time frame
of measurements are very long, the procedures or instrument are different or the operator
has changed.
259
Paper F
Inter-instrument agreement describes the reproducibility of two or more instruments of

identical design (In this study e.g. instrument 1-3 have identical design). Inter-model
agreement expresses the reproducibility of two or more instruments of different design
(e.g. instrument 4 and instrument 5). In other words, reproducibility determines the
variations between instrument’s readings. Instruments, which have the same design the
random amount of bias is reduced compare to instrument with different design. The two
types of reproducibility can be tested in a similar way. The most common way of testing is
pairwise color difference assessment of a series of specimens. Various color difference
parameters are used in the literature [10] including the mean color difference, maximum,
the root mean square (RMS) color difference or the MCDM. RMS color difference ΔE is
calculated as
, (2)
which is similar to average standard deviation, for N values of ΔE.
3. Experimental approach
3.1 Methods and procedures

To determine the accuracy of the used instruments the color difference has been calculated
between the measurement average value and the corresponding ‘true value’. For each
instrument 15 measurements of each tile has been taken in a sequence and the average has
been calculated. In this work, the ‘true value’ is related to the reference values provided by
CERAM who is a high-accuracy laboratory and the manufacturer of the used 14 BCRA
tiles. Due to practical reasons only the short-term repeatability including 15 measurements
in sequence and the long-term repeatability with 10 weeks interval of the instruments have
been assessed. The sample used for this test is the White BCRA tile and the color
differences were calculated according to MCDM. It is worth mentioning that before the 15
actual measurements have been conducted to assess the short-term repeatability and after
10 weeks to determine the long-term repeatability a warm up procedure including 25
measurements in a row on its own white standard has been performed. To determine the
inter-instrument agreement 14 BCRA tiles have been quanified and the mean from each
set has been calculated and consequently compared instrument pairwise. To assess the
inter-model agreement the color difference between the average of 15 measurements of
260
Paper F
each BCRA sample has been calculated. Thus, the pairwise contrast was compared
between the different types of instruments.
3.2 Instruments
In this paper, nine commercial spectrophotometers (one bench-top and 8 hand-held)
typically used in the graphic arts industry have been analysed. Table 1 presents the
instruments and their manufactor’s specifications. Because one of the manufacturers is
requesting not to publish their name the names were anonymised by identifying the
instruments with instrument 1, instrument 2… instrument 9.
All instruments represent bidirectional measurement geometries and uses similar light
sources. The instrument’s aperture size varies from 2 mm up to 4,5 mm. Note, that only
two instruments had valid certification (VC) at the time of testing and for seven
instruments the manufactory’s certification have expired (EC). Nevertheless, the authors
were aware of the situation obtaining measurements from 7 instruments, which were not
re-certified at the time of the experiment. The 7 (uncertified) instruments are between three
and eight years old and not used very often. However, due to the fact that the presented
study is intended to demonstrate the real situation and not reproducing laboratory
conditions the choice of the used instruments can be justified. According to the authors
experiences from the field it is not uncommon to use instruments in the printing industry,
which the re-certification have expired. Therefore the instruments used in this study
represent the realistic situation, which is common in the practical production environment.
As stated in ICC [11] when comparing, instruments can be divided into product families
which are instruments of the same model from the same manufacturer using equal
parameters (e.g. in this work instrument 1-3 can be considered as one family and named
‘Family A’ and instrument 6-9 ‘Family B’). In terms of repeatability, or reproducibility,
instruments with identical design (inter-instrument) or different design (inter-model) can
also be compared.
261
Paper F
Mea- Aper- Mea- Light Manufac- Spectral Manufac-

suring ture suring source turer’s range turer’s
without in geo- claimed and claimed
replace- mm metry precision interval Short term
ment repeatability
Instrument 1 Mean ΔE*ab 0.02
(EC) Family A Tungsten ΔE*ab 0.3 (Standard
45°x:0°
lamp, gas Max 380nm to shift from 10
Instrument 2 Yes 4 filled, ΔE*ab 0.8 730nm at measurements
(EC) type A by 12 10nm at 10 sec.
Instrument 3 light BCRA interval on
(EC) tiles white)
Mean
ΔE*94 <
45°x:0°
380nm to
Instrument 4 1.0 ΔE*94 < 0.2
No 2 Type A 780nm at
(VC) by 12
10nm
BCRA
tiles
ΔE*ab 0.02
Tungsten Mean (Standard
45°x:0°
lamp, gas ΔE*ab 0.3 380nm to shift from 10

Instrument 5 Yes
4.5 filled, by 12 780nm at measurements
(EC)
type A BCRA 10nm at 10 sec.
light tiles interval on
white)
Instrument 6 ΔE*94 < 0.1
Mean
(EC) (From 10
Tungsten ΔE*94 0.4
Instrument 7 measurements
Family B
45°x:0°
lamp, gas Max 380nm to

(EC) at 3 sec.
4.5 filled, ΔE*94 1.0 780nm at
Instrument 8 Yes interval on
type A by 12 10nm
(EC) white)
light BCRA
Instrument 9 tiles
(VC)
Table 1. Overview of the nine instruments (two with valid certification [VC] and seven
with expired certification [EC]) used in this work and the corresponding manufacturer’s
specifications.
Different components (such as light source, detector and dispersing element) and their
properties consisting of the measurement instrument determine the value of a measured
sample. However, it is not the scope of that work to investigate these properties and their
contribution to the measurement results. A comprehensive overview of color measurement
fundamentals and different instrument components is given by Battle [12].
3.3 Calibration procedure

Before conducting the measurements, normal warm up and calibration procedures were
followed. To warm up the instrument, 25 measurements in a row on its own white standard
were made. Consequently, each instrument has been calibrated on its own white reference
262
Paper F
tile supplied by the manufacturer along with the instrument (absolute white calibration).
The absolute spectrum of the white reference tile is stored in the instrument and during the
calibration the obtaining spectral response is adjusted so that it matches the stored
spectrum. Thus, the instrument’s internal software is calculating the spectral reflectance of
the measured samples [13]. Instrument manufacturers typically recommend a calibration
procedure at least once a day or if numerous measurements are undertaken in rapid
succession. Furthermore, if thermal oscillation occur perhaps due to changes in the room
temperature or due to the instrument’s measurement lamp which can be frequently
switched on recalibration is recommended to keep the measurements constant [14].
3.4 Test procedure on BCRA tiles

To evaluate the performance of the instruments in terms of accuracy, repeatability and
reproducibility a series of British Ceramic Research Association (BCRA) Ceramic Color
Standards Series II (CCS II) ceramic tiles have been employed [15]. In this paper, 14
BCRA ceramic gloss tiles including one Black and one White BCRA ceramic gloss tile
were measured and compared to the ‘true value’ of the BCRA tiles to determine the
instruments performance. The measurement procedures were done according to ISO 13655
[16].
3.5 Test procedure on printed substrates

In order to analyse the measurement uncertainities of the color measurement instruments
on commercial printed substrates, measurements were conducted on the UGRA/FOGRA
Media Wedge CMYK [17], which includes 46 color patches (Figure 4). The Media Wedge
was printed on three different paper substrates. The first paper substrate was a hard-copy
digital proof print, printed according to the ISO 12647-7 graphic art standards for paper
type 1 simulation by a commercial printing house. The second paper substrate was paper
type 1 printed by the same commercial printing house aiming the ISO 12647-2 graphic art
standards. And the third paper substrate was paper type 5 Altona testsuite reference print
[18].
Before measuring the color patches of the Media Wedge, warm up and calibration
procedure was followed as discussed previously. The Media Wedge was measured three
times in a sequence with each instrument. White backing material in accordance with ISO
13655 [16] was used.
263
Paper F
Figure 4. Figure 14, UGRA/FOGRA Media Wedge CMYK.
3.6 Data collections

All instruments used in this paper reported spectral reflectance factor values from 380nm
to 730nm with 10nm interval. Spectral measurements taken from the BCRA tiles and the
paper prints were converted to CIEXYZ tristimulus values according to the CIE 1931 2°
observer and the CIE Standard illuminant D50 using the method proposed by ASTM 308,
Table 1 [19]. Further details on these calculations are well documentet in e.g. Hunt [20]
too. Furthermore, CIELAB (D50 as the reference white) values were calculated according
to CIE 15 specifications. Colorimetric difference ΔE*ab and ΔE*94 (as some manufacturers
quoted the color difference in ΔE*94) values were computed between the BCRA reference
data and the measurements data obtained using each instrument [21]. All measurements in
this study have been conducted in the same location and the identical room temperature
conditions.
4. Experimental results and discussion
As mentioned before the presented article is aiming to evaluate the performance of color
measurement instruments in terms of precision and accuracy and is demonstrating the
effect of color measurement variability in a color managed printing workflow.
4.1 Measurement accuracy

As stated earlier ISO defines accuracy as the conformance of a series of measurements to
the accepted value for a given sample. In other words how closely an instrument can
conform to a certain reference or ‘true values’. In this work the reference values have been
provided by CERAM who is the manufacturer of the used 14 BCRA tiles. Figure 5 shows
the color difference between each instrument’s reading on the BCRA tile and the
corresponding ‘true value’. Overall, it can be observed that the chromatic BCRA tiles (e.g.
Red, Orange and Yellow) produces larger color differences than the achromatic BCRA
tiles. Furthermore, all the instruments demonstrate a smaller color differences for the Black
tile than the White tile and the dark to light grey tiles show very similar behavior.
264
Paper F
Figure 5. Color difference of nine instruments according to the 14 BCRA tiles reference.
The better performance of the instruments on the Black tile suggests good black trap
calibration. On the other hand the larger differences on the White tile may indicate that the
instruments do not agree very well on the definition of white, which could be traced to the
instruments’ calibration or missing certification (Table 2).
Black White
Instrument 1 (EC) 0.35 0.71
Family A Instrument 2 (EC) 0.16 0.22
Instrument 4 (VC) 0.13 0.87
Family B
Instrument 9 (VC) 0.18 0.35
Table 2. Color difference ΔE*ab results of nine instruments according to the Black and
White BCRA tiles reference.
According to the White tile results shown in Table 2 the instruments of the Family A range
from 0.16 to 0.48 ΔE*ab, for the instruments of Family B the differences are even larger,
0.18 to 1.11 ΔE*ab. This again confirms that some of the instruments didn’t comply within
the calibration on white reference tile supplied by the manufacturer or this could be a
consequence of out of date re-certification. On the other hand, the newest and certified
265
Paper F
instrument (instrument 4) shows a rather high color difference on the White tile (0.87
ΔE*ab) whereas the instrument 8 performs the least color difference on the White tile
although, the re-certification has been expired a long time ago. Similar results with rather
small color difference can be seen from instrument 2 (0.22 ΔE*ab). According to the
presented results on the White tile, there is no significant evidence whether an expired
certification effect the measurement results. Questions like how frequently an instrument is
used and how an instrument is treated and maintained by the operator determine the
precision of the instrument. However, the authors recommend an appropriate maintenance
of the instrument including regular instrument re-certification to approve the obtained
measurements.
The poorest performance for almost all instruments results from the measurements of the
Red, Orange and Yellow tiles as shown in Figure 6. Except for the instrument 4 the
Orange tile produces the largest color difference. Instrument 7 shows the least color
differences for these three tiles. Considering the color differences within the product
families (instrument family A includes instruments 1-3, and family B includes instruments
6-9) on the tiles Red, Orange and Yellow, there is no obvious trend visible.
Figure 6. Color difference of nine instruments according to the BCRA tiles Red, Orange
and Yellow reference.
Note, it is important to consider the inherent physical properties of the BCRA tiles.
According to a previous work by Fairchild and Grum [22] they stated that the BCRA tiles
Red, Orange and Yellow can exhibit appreciable thermochromism due to sharp changes in
their spectral reflectance curves. Therefore, based on this findings a study by Berns [23]
proposed against using the tiles Red, Orange and Yellow unless the temperature of the tiles
at the time of calibration was known and this temperature was maintained both at the
location where the tiles would be used and during their measurements. However, according
266
Paper F
to the results shown in Figure 6 there is no clear evidence of thermochromism for the
instrument 4 and instrument 7 except for the yellow tile. In contrast, the ‘master
instrument’ demonstrates larger color differences due to possibly generating significant
heat in the measuring process. According to Fairchild and Grum [22], it is important to
make sure that the temperature of calibration standards remains constant during their use.
On the other hand, there have no significant color changes be observed with small
temperature changes around room temperature.
Another way of examining the measurement distribution is to assess the dispersion of the
measurements on the CIELAB a*- b*plane. Figure 7 illustrates the measurements of nine
instruments on the Orange tile including the distance to the reference itself.
Although the results of all measurements show a rather low accuracy, a relatively high
precision of the instruments can be considered due to the measurement dispersion, which
lies almost in one quadrant in the CIELAB system. Figure 8 shows the measurement value
distribution of all instruments on the Orange tile including reference displayed on CIELAB
L*, C* plane. It can be seen that except for instrument 4, the C* color differences
comparing to the reference can be considered as rather large. On the other hand, the L*
differences can be considered as low. Addressing the product families, it can be noticed
that instrument 4 performes best on Orange comparing to the other instruments with
different designs.
Figure 7. Measurements of nine instruments on BCRA tile Orange including reference

displayed on CIELAB a*, b*plane.
267
Paper F
Figure 8. Measurements of nine instruments on BCRA tile Orange including the reference
displayed on CIELAB L*, C*plane.
Figure 9 illustrates the spectral reflectance of all nine instruments measured on the Orange
tile. It is also clearly noticeable that in the area between 600nm and 730nm the dispersion
of the measured spectral reflectance by the instrument is rather large. It can be only
speculated what the reasons can be for the large dispersion. One reasonable explanation
can be some degree of thermochromism in combination with some white point error due to
expired instrument certification. In addition, the reflectance factors of most of the
instruments are far below the reference reflectance factor. Finally, it can be noticed that
instrument 4 shows the closest spectral reflectance curve to the reference and again
confirms the least color difference on the Orange tile as seen in Figure 6.
268
Paper F
Figure 9. Spectral reflectance measurements of nine instruments on BCRA tile Orange

including the reference.
4.2 Short-term and long-term repeatability

Table 3 shows the manufacturer’s agreement and the corresponding results in terms of the
short-term and long-term measurements. Note, for the long-term repeatability evaluation
instrument 1, instrument 6 and instrument 7 were not accessable. Although the
manufacturers do not specify any particular measurement agreements for the long-term
repeatability it might be obvious that it is the degree to which the instrument makes
identical measurements over a long time.
Instruments Manufacturer’s agreement Short-term Long-term

repeatability repeatability
Instrument 1 (EC) Fail n.a.
ΔE*ab 0.02 (standard shift from 10
Instrument 2 (EC) Pass Pass
measurements at 10 sec. interval on white)
Instrument 3 (EC) Pass Pass
Instrument 4 (VC) ΔE*94 < 0.2 Pass Pass
Instrument 5 (EC) ΔE*ab 0.02 (standard shift from 10
Fail Pass
measurements at 10 sec. interval on white)
Instrument 6 (EC) Pass n.a.
Instrument 7 (EC) ΔE*94 < 0.1 from 10 measurements at 3 Pass n.a
Instrument 8 (EC) sec. interval on white) Pass Pass
Instrument 9 (VC) Pass Pass
Table 3. Overview over short-term and long-term repeatability performance.
269
Paper F
Except for the instrument 1 and instrument 5 on short-term repeatability, all instruments
perform results, which qualifies them to pass for the short-term repeatability according to
the manufacturer’s agreement. Note, that the certification for the instrument 1 and
instrument 5 has expired. On the other hand, all available instruments have passed the
long-term repeatability test.
Figure 10 shows the performance of the short-term and long-term repeatability of the
instrument 9 and the manufacturer’s agreement which is defined with ΔE*94 0.1 with
respect to the mean CIELAB value of 10 measurements on white. The x-axis indicates the
short-term and long-term repeatability variations whereas on the y-axis the color difference
is represented. The closer the horizontal mean-lines (Oct08 E94 Mean and Jan09 E94
Mean) are, the more identical are measurements and hence better the long-term
performance can be considered.
Figure 10. Short-term and long-term repeatability on white including manufacturer’s

agreement ΔE*94 0.1 for instrument 9.
It can be seen that both the short and long-term repeatability performs almost equally and
within the manufacturer’s agreement. Furthermore, the graph shows that the largest
variations are in the beginning of the measurement sequence. Hence, increasing the
number of measurements in the warm up time procedure would increase the total
performance of the repeatability for this instrument. It is worth mentioning that instrument
9 has been recently re-certified by the manufacture. Therefore it can be speculated that the
recertification of the instrument can be the reason for the excellent performance of the
short-term and long-term repeatability.
For the instrument 5 the manufacturer reduces the short-term repeatability to ΔE*ab 0.02
units. In Figure 11 it can be clearly observed some minor short-term measurement
variations and that the overall repeatability measurements for the instrument 5 exceed the
270
Paper F
manufacturer’s agreement. Hence, instrument 5 does not conform to the manufacturers’

specifications and need to be re-certified. On the other hand, although the manufacturer is
not providing any long-term repeatability specification the long-term repeatability can be
considered as very good due to the almost identical measurements between the 10 weeks
interval.
Figure 11. Short and long-term repeatability on white including manufacturer’s agreement
ΔE*ab 0.02 for the instrument 5.
In our paper, the manufacturer of the instrument 4 has defined the largest repeatability
agreement with ΔE*94 of 0.2. Although the mean measurements are strongly inside the
manufacturer’s agreement as can be seen in Figure 12, the short-term measurement
variations are rather large. The variation is very apparent regardless that instrument 4 is
new and recently certified by the manufacturer. On the other hand, the long-term
repeatability illustrates almost the same variations. Hence, the long-term repeatability can
be considered as acceptable. However, comparing the short-term and long-term
repeatability performance with another instrument family the variations are rather large,
e.g. the short and long-term variation of the instrument 4 is much larger then the
manufacturer’s short-term agreement for e.g. the instrument 5. Figure 13 demonstrates the
L* versus the measurement number for the three instruments 4, 5 and 9. Although,
instrument 5 and instrument 9 show reasonable repeatability performance in L*, the drift in
instrument 4 is very apparent, especially when considering the very short time scale of the
measurements. It can be speculated in which direction the drift would have continued by
increasing the number of measurements. The rather large variations might be explained
due to instrument technology and the user interface of the instrument 4. The short-term
measurements have been performed with replacement/UpDown settings, which means that
the tile is not left in place at the instrument’s aperture when measuring. Furthermore, it has
271
Paper F
to be noted that the aperture of this instrument is very small, only 2 mm in diameter. The
aperture size in combination with the physical measurement settings,
replacement/UpDown may contribute to the obtained measurement variations.
Figure 12. Short and long-term repeatability on white including manufacturer’s agreement
ΔE*94 0.2 for the instrument 4.
Figure 13. L* versus measurement number for instrument 4, instrument 5 and

instrument 9.
4.3 Inter-instrument agreement

The instrument manufacturers define certain inter-instrument agreements within their
instrument families [24]. For the instrument family B (instruments 6-9) the manufacturer
272
Paper F
has defined an inter-instrument agreement of mean ΔE*94 of 0.4 and Max ΔE*94 of 1.0 for
single measurement mode on 12 BCRA tiles (D50, 2°). Figure 14 demonstrates the
pairwise contrast of the inter-instrument agreement within the instrument family B. It can
be seen that instrument 6, 8 and 9 meet the manufacturer’s requirements both in terms of
mean ΔE*94 < 0.4 and Max ΔE*94 < 1.0. On the other hand for the instrument family A the
manufacturer has defined an inter-instrument agreement of mean ΔE*ab 0.3 and Max ΔE*ab
0.8 for single measurement mode on 12 BCRA tiles (D50, 2°). Figure 15 shows the
pairwise contrast of the inter-instrument agreement within the instrument family A
(instrument 1-3). It can be seen that even though the direct comparison between instrument
2 and instrument 3 is within the inter-instrument agreement given by the manufacturer
regarding max and mean ΔE*ab have been slightly exceeded. The pairwise comparisons
between instrument 1 and instrument 2, and between instrument 1 and instrument 3 exceed
the manufacturer’s requirements distinctly in terms of mean and max ΔE*ab. Instrument 7
exceeds the requirements noticeably in both the mean value and Max value.
Figure 14. Pairwise contrast of the inter-instrument agreement within the instrument
family B, compared to the manufacturer’s specifications.
273
Paper F
Figure 15. Pairwise contrast of the inter-instrument agreement within the instrument
family A, compared to the manufacturer’s specifications.
4.4 Inter-model agreement

In order to determine the performance of the inter-model agreement instruments from
different families have been compared (instrument 2, instrument 4, instrument 5 and
instrument 9). Table 4 shows one color difference ΔE*ab for each 14 BCRA sample for
each pair of instruments. From each set of color differences, the mean, maximum and RMS
color differences and representing standard deviation have been computed. The pairwise
contrast where instrument 4 is involved is very apparent with maximum color differences
between ΔE*ab 2.76 and ΔE*ab 4.46 in the BCRA tiles Red, Orange and Yellow which
again is not surprising having seen the accuracy results previously. On the other hand,
looking at the pairwise contrast between instrument 5 and instrument 9 the inter-model
agreement can be considered as rather good with a maximum ΔE*ab < 1. In a previous
work conducted by Wyble [9] the RMS results from a very similar test using three
bidirectional instruments show significant larger color differences. Again, the tiles Red,
Orange and Yellow were responsible for the largest color differences. Although, the
instrument models are unknown, it can be assumed that the difference between the
instrument design was larger compared to the instrument design presented in the present
work.
The dark to light grey tiles show very similar behavior as already seen previously in the
results of the accuracy. Generally, it can be observed that dark achromatic tiles result in a
significant better pairwise instrument performance than measurement from chromatic tiles
with respect to the inter-model agreement.
274
Paper F
According to ASTM E2214 [10] the difference between inter-instrument and inter-model
agreement can be as large as an order of magnitude which can be confirmed by the
presented results with certain pairwise instrument combinations. Note, filter based
colorimeter providing tristimulus values only were not part of this study. It can be assumed
that a pairwise comparison between instruments obtaining spectral data versus tristimulus
data the inter-model agreement can be rather poor. Previously, reports by Rich et al. [25]
have reported rather large color differences considering inter-instrument reproducibility.
Furthermore, in his article, the inter-model agreement between the colorimeters and
spectrophotometers used for emission measurement has shown a very large color
difference.
2 vs 4 2 vs 5 2 vs 9 4 vs 5 4 vs 9 5 vs 9
Pale grey 0.68 0.41 0.16 0.37 0.79 0.56

Mid Grey 0.75 0.31 0.07 0.46 0.79 0.36
Diff Grey 0.79 0.50 0.27 0.38 0.58 0.41
Deep Grey 0.26 0.17 0.28 0.32 0.03 0.33
Deep Pink 1.10 0.46 0.43 0.65 0.99 0.53
Red 2.76 1.48 1.07 1.31 1.73 0.45
Orange 4.27 1.76 1.46 2.68 2.97 0.31
Yellow 4.46 1.44 1.06 3.27 3.82 0.71
Green 0.85 0.31 0.66 0.84 1.25 0.54
Diff Green 1.01 0.56 0.53 1.22 1.49 0.51
Cyan 2.12 0.81 0.26 1.36 1.87 0.61
Deep Blue 0.56 0.06 0.93 0.52 1.38 0.93
Black 0.28 0.18 0.27 0.12 0.14 0.20
White 0.80 0.58 0.30 0.25 0.75 0.52
MEAN 1.48 0.64 0.55 0.98 1.33 0.50

MAX 4.46 1.76 1.46 3.27 3.82 0.93
RMS 2.00 0.83 0.69 1.34 1.66 0.53
STDEV 1.40 0.54 0.42 0.94 1.04 0.18
Table 4. Pairwise contrast of instruments using BCRA tiles
275
Paper F
4.5 Results of print measurements

The following are the results from the measurements performed with seven instruments on
three types of substrates (instrument 6 and instrument 7 were not accessible in this task of
the work). Firstly, the measurement results on substrate proof will be presented followed
by the results for paper type 1 and paper type 5 respectively. To recap, the proof has been
created in a commercial printing house, simulating ISO 12647-2 paper type 1. 46 color
patches of the UGRA/FOGRA Media Wedge CMYK [17] have been measured three times
in sequence and consequently the average were calculated.
The mean value (of the three measurements per patch) have been used to calculate the
color difference between the target values and each single instrument. The CIELAB target
values of the UGRA/FOGRA Media Wedge CMYK are based on print conditions as stated
in ISO 12647-2 and the appropriate characterisation tables for different paper types are
provided by Fogra [26]. Table 5 shows the calculated color difference values compared
with the CIELAB ΔE*ab tolerances according to ISO 12647-7. It can be seen that five
instruments (instrument 1, instrument 2, instrument 3, instrument 5 and instrument 8) have
performed measurements, which are within the acceptable tolerances. The measurements
of the instrument instrument 4 and instrument 9 show results, which are far outside the
defined tolerances. At first glance, the verdict might be justified.
276
Paper F
Sub- Mean Max Primaries Composed

strate grey
ΔE*ab 3 ΔE*ab 3 ΔE*ab 6 ΔE*ab 5 ΔH* ΔH*
(ΔE*ab 2,5) (ΔE*ab
1,5)
C M Y K C M Y Average
Instrument 1 1.36 1.39 2.65 1.2 1.52 1.81 1.23 0.76 0.81 0.4 0.48
Instrument 2 1.69 1.28 3.0 0.9 1.51 0.66 1.2 0.48 1.36 0.04 1.08
Instrument 3 1.52 1.4 3.12 1.43 1.7 0.9 1.38 1 1.59 0.45 0.93
Instrument 5 0.92 1.26 2.46 0.87 1.17 2.05 1.04 0.26 0.86 0.15 0.71
Instrument 8 1.4 1.12 2.67 0.66 1.07 1.48 1.1 0.31 0.92 0.06 0.71
Instrument 9 6.34 3.04 6.34 3.27 2.36 2.49 1.68 3.14 2.19 0.33 3.47
UV cut
Instrument 4 1.4 2.54 7.5 2.96 3.03 7.5 1.36 2.56 0.05 0.71 0.55
Table 5. Color differences on proof of seven instruments including the CIELAB ΔE*ab
tolerances according to ISO 12647-7:2007 (Orange marked values are outside the
tolerance).
Although instrument 4 performs satisfactorily for most of the colors, the color difference
between the instrument’s measurement and the reference on the primary color yellow is
ΔE*ab > 7, which is a considerably large color difference. Instrument 9 is the only device,
which is using an UV cut filter. Therefore it is obvious that the measurement on the proof
substrate exceeds the tolerance due to the concentration of optical brighteners which
effects the CIE b* value most (from reference b* -2 to measured b* +4).
Looking at the above measured values, if the proof would have been measured initially in
the print shop (where the proof is generated) with e.g. the instrument 1 and then measured
by the customer with e.g. the instrument 4 or instrument 9 (which contains the UV Cut
filter), then, only the measurement performed by the instrument 1 would have been
considered as within the tolerance. However, the customer would not have accepted the
proof as approved in a first attempt, as the measurements made by his instrument exceed
the tolerances. Using an instrument with a UV Cut filter and an instrument without
measuring the proof is inappropriate can be considered as an obvious operator error.
It has been observed previously that the instrument 4 results in a large color difference in
the primary color yellow when compared with the reference. Figure 16, which shows the
277
Paper F
measurements of seven instruments on proof substrate on the primary color yellow

including reference displayed on CIELAB a*, b* plane can confirm this finding. However,
looking at the precision of the other instruments, the graph illustrates a very small
dispersion of the measurements taken. Furthermore, the instrument family A (instrument 1,
instrument 2 and instrument 3) can clearly be recognised as the one with the highest
precison. Instrument 9, on the other hand, shows a larger difference in the CIE b* value as
seen earlier due to the concentration of optical brighteners in the proof substrate and the
measurement with a UV cut filter. Therefore, this large difference can’t be considered as a
systematic error.
Figure 16. Measurements of seven instruments on proof substrate primary color yellow
including reference displayed on CIELAB a*, b*plane.
Figure 17. Measurements of seven instruments on proof substrate primary color yellow
including reference displayed on CIELAB L*, C*plane.
Looking at the measurement results on CIELAB L*, C* plane (Figure 17) the precision
within the instrument families can be considered as good. Although, the difference in L*
278
Paper F
value between the instrument 4 and the other instrument families is rather small, the
difference in C* value can be recognised as very large.
Figure 18. Measurements of seven instruments on proof substrate primary color cyan
including reference displayed on CIELAB a*, b*plane.
Similar measurement patterns can be observed in other primary (cyan and magenta) and
secondary colors (red, green and blue). Figure 18 shows the measurements of seven
instruments on proof substrate on the primary color cyan including reference displayed on
CIELAB a*, b*plane. Overall, it can be observed that the color difference in CIE b* is
larger than on CIE a*. It is apparent that the concentration of the optical brighteners in the
proof substrate again has effected the cyan measurement with instrument 9. Although the
dispersion of the measurements within the instrument families is slightly larger compared
to the primary color yellow, the variations can still be considered as acceptable. It has to be
emphasised that in this task the dispersion of the instrument’s measurements should be
taken into account and not the color difference between the instrument measurements and
the reference.
Another way of assessing the measurement results on the solid primary colors is by
comparing the inter-model agreement. The observation made in Figure 18 can be
confirmed with the CIE ΔE*ab values in Table 6 which shows the color differences ΔE*ab
on the solid cyan and magenta between each instrument. It is apparent that the instrument 4
and instrument 9 result in the largest color differences on cyan when compared with the
other instruments (e.g. Color difference of ΔE*ab 6.12 between instrument 4 and
instrument 9). This rather poor inter-model performance has already been observed in
Table 4 by the pairwise comparison of the two instruments e.g. on BCRA tile Cyan. The
279
Paper F
differences between instrument 4 and the other instruments range between ΔE*ab 2.27 and
4.09. On the other hand, the inter-instrument performance between the instrument family A
(instrument 1, instrument 2 and instrument 3) can be considered as acceptable with
differences ranging between ΔE*ab 0.6 and 1.88. Looking at the results of the instrument 5
and instrument 8 the ΔE*ab is less than 0.8.
The upper triangle in Table 6 shows the results on solid magenta, where again, the
instrument pair instrument 4 and instrument 9 show the largest ΔE*ab of 4.16. The least
color differences are not within the instrument family A (instrument 1, instrument 2 and
instrument 3) itself but between instrument 8 and instrument 5 (ΔE*ab < 0.5) and
instrument 8 and instrument family B (ΔE*ab < 0.9).
Table 7 shows the color differences ΔE*ab on the solid yellow and black. Regarding
measurements on yellow again, the instrument 4 shows the most significant color
differences compaired to the other devices with ΔE*ab > 5.12 which, already has been seen
on CIELAB a*, b*plane in Figure 16. On the other hand, instrument 9 shows a much
better precison on yellow than what we have seen on the color cyan and magenta.
The instruments performance on solid black, however, shows measurement results, which
are almost ΔE*ab < 1.0 across all instrument combinations including instrument 4 and
instrument 9. A very similar measurment performance on the BCRA tile black has been
observed previously in Table 2. Hence, black seems to be the least critical color
considering the precision on inter-instrument and inter-model agreement.
Table 6. Inter-instrument agreement on proof substrate in solid cyan (lower left half of the
table) and magenta (upper right half of the table) between all instruments.
Table 7. Inter-instrument agreement on proof substrate in solid yellow (lower left) and
black (upper right) between all instruments.
280
Paper F
Below the results on substrate paper type 1 according to the ISO 12647-2 standard will be
presented. The mean values (of the three measurements) have been used to calculate the
color difference between the reference given by ISO 12647-2 paper type 1 (white backing)
and each single instrument. The calculated color difference have been compared with the
CIELAB ΔE*ab tolerances according to ISO 12647-2.
It can be seen in Table 8 that only three instruments (instrument 2, instrument 5 and
instrument 8) give measurement results, which will qualify the print as approved due to the
values obtained, which are within the printing ISO tolerance values for all primary colors
and the substrate.
There is evidence of optical brighteners being present in the paper type 1 substrate which
affects the CIE b* value when measuring with instrument 9. Therefore using instrument 9
will exceed the measurement value of the substrate above the tolerance value (ΔE b*±2).
Moreover, substrates containing optical brighteners affect not only the measurement results
on the substrate but also colors in the blue regions when measuring with instrument
including UV cut filter. For that reason, the cyan measurement is rather high too. This
observation has been confirmed previously by a work conducted by Radencic [27] where
he concluded that colors which produce extremely high color differences regardless of the
instrument were generally recorded on the substrates containing optical brightener and
were generally blue in shade. Instrument 1 and instrument 3 show measurement values on
black, which just exceeds the color differences tolerances too, as well as instrument 4 in
yellow.
Substrate Primaries
ΔE L*±3 ΔE a*±2 ΔE b*±2 ΔE*ab 5
C M Y K
Instrument 1 0.3 1.25 1.41 4.24 1.2 1.86 5.24
Instrument 2 0.07 1.19 0.91 4.55 0.68 2.48 4.8
Instrument 3 0.04 1.42 1.19 4.91 1.05 2.83 5.23
Instrument 5 0.05 1.1 1.41 4.1 1.16 1.89 4.52
Instrument 8 0.15 1.3 1.2 4.16 0.57 2.41 4.02
Instrument 9 0.38 0 3.6 5.35 2.11 2.67 4.0
Instrument 4 0.5 1.79 1.35 3.89 4.07 5.01 4.37
Table 8. Color differences on substrate paper type 1 of seven instruments including the
CIELAB ΔE*ab tolerances according to ISO 12647-2 (Orange marked values are outside
the tolerance).
281
Paper F
Figure 19 shows the measurements of seven instruments on substrate paper type 1 on the
primary color cyan including reference on CIELAB a* - b*plane. Overall, a very similar
pattern considering the dispersion of the measurement as seen previously on proof can be
observed. Also, the color difference in CIE b* is larger then on CIE a*.
Figure 19. Measurements of seven instruments on substrate paper type 1 primary color
cyan including reference displayed on CIELAB a*, b*plane.
Table 9 shows the color differences ΔE*ab on the solid cyan and magenta between each
instrument on substrate paper type 1. It can be recognised that the inter-instrument and
inter-model agreement on substrate paper type 1 is almost identical to the inter-instrument
and inter-model agreement on substrate proof. The same can be stated for solid yellow and
solid black for paper type 1 as seen in Table 10.
Table 9. Inter-instrument agreement on substrate paper type 1 in solid cyan and magenta
between all instruments.
Table 10. Inter-instrument agreement on substrate paper type 1 in solid yellow and black
between all instruments.
282
Paper F
And finally the results on substrate paper type 5 according to the ISO 12647-2 standard
will be presented. Table 11 shows the color difference results on substrate paper type 5 of
seven instruments. It can be seen that all instruments performed measurements, which will
qualify the print as approved due to the values obtained, which are within the printing ISO
tolerance values for all primary colors. Although this paper type 5 should not contain any
concentration of optical brighteners (as stated by the paper manufacturer) instrument 9
(UV cut) shows a CIE b* value (2.8) which exceeds just the given tolerance. Instrument 5
gives the closest readings compare to printing ISO tolerance values.
Figure 20 shows the measurements of seven instruments on substrate paper type 5 on the
primary color cyan including reference on CIELAB a* - b* plane. The dispersion of the
measurements is almost identical again with the dispersion of measurements seen on
substrate proof and substrate paper type 1.
The inter-instrument and inter-model agreements on the solid primary colors cyan,
magenta, yellow and black on substrate paper type 5 are almost identical to the inter-
instrument and inter-model agreement on substrate proof and paper type 1 respectively.
Substrate Primaries
ΔE L*±3 ΔE a*±2 ΔE b*±2 ΔE*ab 5
C M Y K
Instrument 1 0.5 0.3 1.7 2.04 3.5 1.44 2.27
Instrument 2 0.0 0.3 1.9 2.65 2.53 1.22 1.55
Instrument 3 0.0 0.4 1.8 2.78 2.18 1.47 1.55
Instrument 5 0.0 0.2 1.0 1.67 2.14 1.68 1.29
Instrument 8 0.0 0.32 2.0 2.44 3.22 1.02 1.47
Instrument 9 0.2 0.25 2.8 2.61 3.76 1.26 1.56
Instrument 4 0.2 0.12 0.8 2.33 3.51 3.72 1.28
Table 11. Color differences on substrate paper type 5 of seven instruments including the
CIELAB ΔE*ab tolerances according to ISO 12647-2 (Orange marked values are outside
the tolerance).
283
Paper F
Figure 20. Measurements of seven instruments on substrate paper type 5 primary color
cyan including reference displayed on CIELAB a*, b*plane.
5. Conclusions
Nine commercial spectrophotometers typically used in the graphic art industry were
evaluated in terms of precision and accuracy and the effect of color measurement
variability was demonstrated. As stated in ASTM E2214 [10] the most important
specification in color measuring instrument is repeatability. According to our results,
except of two instruments all others did pass the manufacturer’s agreement in the short-
term repeatability test. All available instruments did conform to the long-term agreement.
On the other hand, as discussed above, there are large differences between inter-instrument
and inter-model reproducibility. On one side, the obtained results from the inter-instrument
test have demonstrated a reasonable performance among instruments within the same
family, as agreed by the manufactures. Nonetheless, the inter-model measurement results
have shown, as expected, much larger color differences, especially using instruments from
different manufactures. The results of the print measurements did confirm the inter-
instrument and inter-model agreement observed before by measuring the BCRA tiles.
As stated previously accuracy describes the averaging of grouping compared to the centre
of a certain ‘true value’. Considering the accuracy results, overall, it can be observed that
the chromatic BCRA tiles (e.g. Red, Orange and Yellow) produces larger color differences
than the achromatic BCRA tiles perhaps due to possible thermochromism. Furthermore, all
the instruments demonstrate a smaller color differences for the Black tile than the White
tile and the dark to light grey tiles show very similar behaviour. The larger differences on
284
Paper F
the White tile may indicate that the instruments do not agree very well on the definition of
white, which could be traced to the instruments’ calibration or certification status.
However, there has no obvious consistency been observed between certified and non-
certified instruments in terms of their performance on the White tile.
Consciously, instruments with valid certification and instruments with expired certification
have been used in this study to be conforming to the common situation in the printing
workflow. I might be expected that instruments with valid certification perform better than
instrument with expired certification. Although missing instrument re-certification did not
show a significant effect on the measurements it is highly recommended to maintain the
instruments according to the manufactures requirement including appropriate re-
certification procedures.
In conclusion, beside of applying only calibrated and certified instruments a further

obvious consequence will be the use of only one certain instrument family (same model,
same design of instrument from the same manufacturer with the same parameters) in a
color managed printing workflow to preserve reasonable color differences. Finally, prevent
the use of instruments with different filters (e.g. UV-cut filter) in the same workflow to
avoid large errors in measurements. However, in order to improve the colorimetric
performance and inter-instrument and inter-model agreement a method of characterizing
measurement instruments using colorimetric regression technique has to be considered.
It might be of interest to consider other potential directions for further work in the field
color measurement uncertainties. The performance of a number of color measurement
instruments (and measurement technologies), in particular for emission purposes (display)
in terms of precision and accuracy could be evaluated and the possible consequences of the
inter-instrument reproducibility in color managed workflow addressed. Another area
within the standardization process in the graphic art industry is the viewing condition set
up according to the parameters defined in ISO 3664 [28] and ISO 12646 [29]. To perform
the calibration and to verify the appropriate parameters of the ambient light conditions the
same hand-held instruments are used attaching a diffuse light measurement head.
Therefore, the precision and accuracy for ambient light measurements have to be
investigated.
285
Paper F
Acknowledgments
Firstly, we would like to thank Carinna Parraman, Senior Research Fellow, Centre for Fine
Print Research, University of the West of England, Bristol, for providing her instruments
to this project. Secondly, we would like to thank Giorgio Trumpy, Institute of Applied
Physics of the National Research Council, Italy, for providing his instrument.
References
[1] ISO12647-2, Graphic technology – Process control for the production of half-tone
colour separations, proof and production prints – Part 2: Offset printing processes,
ISO, 2004
color separations, proof and production prints – Part 7: Proofing processes
working directly from digital data, ISO, 2007
York, 2000.
[4] ASTM E284-08, Standard Terminology of Appearance, American Society for
Testing and Materials West Conshohocken, PA, USA, 2008
Applied Optics, vol. 8, pp. 775-783, 1969.
Research & Application, vol. 6, pp. 195-202, 1981.
Application, vol. 19, pp. 322-331, 1994.
[8] D. Wyble and D. Rich, "Evaluation of methods for verifying the performance of
vol. 32, pp. 166-175, 2007.
color-measuring instruments. Part II: Inter-instrument reproducibility," Color
of Color Measuring Instruments, American Society for the Testing of Materials,
West Conshohocken, PA, 2008
286
Paper F
[11] ICC. 2008. Precision and Bias of Spectrocolorimeters, . ICC white paper 22.
Available:
http://www.color.org/ICC_white_paper_22_Precision_and_Bias_in_Measurement.
pdf,12. September 2008
[12] D. Battle, "The measurement of colour," in Colour physics for industry, R.
McDonald. 2nd: Society of Dyers and Colourists, 1997, pp. 57-80.
[13] G. Beretta, Spectrophotometer Calibration and Certification vol. 2: HP
Laboratories Palo Alto, 1999.
[14] D. R. Wyble. (2010, 24. August 2010). Color Measurement - Introduction,
Historical Perspective, Definitions and Terminology, Components of a
Spectrophotometer, Light Source, Detector, Dispersing Element. Available:
http://encyclopedia.jrank.org/articles/pages/1246/Color-Measurement.html
[15] BCRA, tiles are produced by CERAM Research, Queens road, Penkhull, Stoke-on-
Trend, ST4 7LQ, England,
[16] ISO13655, Graphic technology – Spectral measurement and colorimetric
computation for graphic arts images ISO, 1996
[17] U. Schmitt and F. Dolezalek, Ugra/FOGRA Media Wedge CMYK V 2.0, Munich,
2004
[18] Print & Media Forum AG, Altona Test Suite Anwendungspaket, Bundesverband
Druck und Medien, 2004
the CIE-System, American Society for Testing and Materials, West Conshohocken,
PA, 2008
[20] R. Hunt, Measuring colour, 3 ed.: Fountain Press, 1998.
[21] CIE15, Colorimetry, CIE Central Bureau, Vienna, 2004
Optics, vol. 24, pp. 3432-3, 1985.
errors," Color Research & Application, vol. 13, pp. 243-256, 1988.
[24] R. Danuser, Standards Laboratory, Personal communication, 2009
Barcelona, Spain 2008, pp. 77-79.
287
Paper F
[26] A. Kraushaar. Characterization data for standardized printing conditions 2008.

Available: http://www.fogra.de/
[27] G. Radencic, "A comparison of spectrodensitometers and their agreement on
common graphic arts materials," in TAGA, San Francisco, CA, 2008.
[28] ISO3664, Graphic technology and photography — Viewing conditions, Geneva:
ISO [www. iso. org], 2009
[29] ISO12646, Graphic technology – Displays for colour proofing – Characteristics
and viewing conditions, ISO, 2008
288
Paper G
Paper G
Peter Nussbaum, Jon Yngve Hardeberg and Fritz Albregtsen
Regression based Characterization of Color Measurement Instruments

in Printing Applications
In Electronic Imaging:
Color Imaging XVI: Displaying, Processing, Hardcopy, and Application,
SPIE Proceedings, 7866,
San Francisco, CA,
2011.
289
Paper G
290
Paper G
Regression based characterization of color measurement instruments in

printing applications
Peter Nussbaum a, Jon Y. Hardeberg a and Fritz Albregtsen b
a
Gjøvik University College, P. O. Box 191, N-2815 Gjøvik, Norway
b
Department of Informatics, University of Oslo, P. O. Box 1080 Blindern, N-0316 Oslo, Norway
E-mail: peter.nussbaum@hig.no, jon.hardeberg@hig.no, fritz@ifi.uio.no
Keywords: Color measurement, color measurement instrument characterization, ISO

standards, print quality, process control, polynomial fitting technique, measurement
uncertainties, Inter-instrument agreement.
ABSTRACT
In the context of print quality and process control colorimetric parameters and tolerance
values are clearly defined. Calibration procedures are well defined for color measurement
instruments in printing workflows. Still, using more than one color measurement
instrument measuring the same color wedge can produce clearly different results due to
random and systematic errors of the instruments. In certain situations where one instrument
gives values which are just inside the given tolerances and another measurement
instrument produces values which exceed the predefined tolerance parameters, the question
arises whether the print or proof is approved or not accepted with regards to the standard
parameters. The aim of this paper was to determine an appropriate model to characterize
color measurement instruments for printing applications in order to improve the
colorimetric performance and hence the inter-instrument agreement. The method proposed
is derived from color image acquisition device characterization methods which have been
applied by performing polynomial regression with a least square technique. Six
commercial color measurement instruments were used for measuring color patches of a
control color wedge on three different types of paper substrates. The characterization
functions were derived using least square polynomial regression, based on the training set
of 14 BCRA tiles colorimetric reference values and the corresponding colorimetric
measurements obtained by the measurement instruments. The derived functions were then
used to correct the colorimetric values of test sets of 46 measurements of the color control
291
Paper G
wedge patches. The corrected measurement results obtained from the applied regression
model was then used as the starting point with which the corrected measurements from
other instruments were compared to find the most appropriate polynomial, which results in
the least color difference. The obtained results demonstrate that the proposed regression
method works remarkably well with a range of different color measurement instruments
used on three types of substrates. Finally, by extending the training set from 14 samples to
38 samples the obtained results clearly indicate that the model is robust.
1. INTRODUCTION
In general, process control is the basic requirement for ensuring satisfactory print and
proof quality in the graphic art industry. To preserve the standardization concept
colorimetric parameters and tolerance values for print and proof productions are defined in
ISO 12647-2 [14] and ISO 12647-7 [15] respectively. Therefore, to ensure the quality
control colorimetric values have to be obtained using color measurement instruments.
Currently, there are many different models of color measurement instruments used in the
printing industry, and this has been found to have significant consequences on print and
proof quality [22].
In a modern color managed and standardized printing workflow, most of the printing
houses use more than one color measurement instrument, typically one instrument in each
department (pre-press, press, and post-press). Moreover, in the context of a Process
Standard Offset (PSO) [21] certification process often the same color control wedge of a
print or proof is measured first by the instrument of the company, that is to be certified,
and secondly with the instrument of the certification body, to determine and confirm
whether the colorimetric values are within the defined ISO tolerances.
However, measuring a control wedge with two different color measurement instruments
will obviously result in different colorimetric data sets due to the nature of the instrument’s
uncertainties [22]. In a certification context assuming that both instruments give values that
are within the given color difference tolerances according to the ISO standard, both
measurements will be approved and the print or proof will be accepted. On the other hand,
if one of the instruments gives values that exceed the tolerances, the question arises which
of the measurement values are correct and which one has failed, even though both
measurement instruments are certified. Depending on the applications and the customer’s
requirements the predefined ISO standard tolerances have been defined narrower to
292
Paper G
increase the print quality. Consequently, the color measurements performed with more
than one instrument are even more critical in terms of the instrument uncertainty.
In the past, a number of studies have addressed the issues of color measurement instrument
accuracy and uncertainties. For more details on assessment of color measuring instruments
in general and inter-instrument reproducibility in particular see the works of Billmeyer [7],
Briggs et al. [9], Billmeyer and Alessi [8], Rodgers et al. [25] and Wyble and Rich [29]. In
a study by Rich et al. [24] the authors have observed that the differences between pairs of
instruments can be quite significant, with maximum differences of up to ΔE*ab of 4.0. In a
previous work by Nussbaum et al. [22] the authors conclude that in order to reduce the
measurement errors in a color managed printing workflow the use of only one instrument
product family (instruments of the same model from the same manufacturer using equal
parameters) is recommended. However, due to a number of different reasons this advice
seems to be rather difficult to implement in the daily printing production environment. A
further technique to reduce the color differences obtained by measuring the same sample
using more than one measurement instrument is applying a correction method to the
obtained color measurements.
Therefore, the aim of the present work is to propose a method to reduce the variations in
color measurement performed with more than one instrument measuring the same color
target. In particular, the main contribution of this study is in characterizing measurement
instruments using a colorimetric regression technique. Finally, the appropriate correction
model applied to the measurement data sets will reduce the color errors between the
measurements obtained by a master instrument and the measurements performed by a
second instrument used (Figure 1). Consequently, the model will improve the colorimetric
performance and inter-instrument and inter-model agreement.
293
Paper G
Un-corrected Corrected
Measurements Correction Measurements
by by
Instrument 1 Instrument 1
Color
Control Before After
Wedge ΔE*ab ΔE*ab
Un-corrected Corrected
Measurements Correction Measurements
by by
Figure 1. A schematic diagram of the color measurement workflow using two measurement
instruments measuring a color control wedge, and applying a correction model to reduce
the errors between instrument 1 and instrument 2.
In order to determine the performance of measurement instruments there are a number of

parameters to consider. According to ASTM E2214 [3] the most important specification is
the repeatability which defines how well an instrument repeats its reading of the same
target over a certain period of time. Reproducibility is a form of repeatability in which one
or more of the measurement parameters have been systematically changed, such that the
target is being different, the time frame of measurements are being very long or the
operator has being changed. Inter-instrument agreement describes the reproducibility of
two or more instruments of the same design and inter-model agreement describes the
reproducibility of two or more instruments of different design. Finally, accuracy describes
the conformance of a series of readings to the accepted or true value. The measurement
variations between instruments can be divided into systematic and random errors.
According to Berns [6], repeatability is affected by random errors including drift,
electronic noise and sample presentation. Random variations are difficult to avoid. On the
other hand the accuracy is affected by systematic errors, which among other characteristics
may be due to different measurement geometry, or detector linearity errors resulting from a
change of wavelength. In the past several attempts have been made to reduce the
systematic errors by characterizing color measurement instruments. The study by Berns [6]
proposes the correction of various systematic errors applying to the spectral measurements
data using multiple linear regression based on modeling the results to improve the
colorimetric performance. In this work, however, the aim is to correct the instrument’s
systematic errors by applying a regression technique directly to the measured CIELAB
data, and hence improve the inter-instrument and inter-model agreement.
294
Paper G
Following this brief introduction including the definition of the aim in this work and
discussing central references, we provide some more methodology information in Section
2, by illustrating the regression model, defining key concepts, and the experimental
procedure including data collection. Then, in Section 3 we present and discuss our results,
before concluding in Section 4.
2. METHODOLOGY
The method we propose in this work is based on color image acquisition device
characterization, which has been applied by implementing polynomial regression with a
least square technique [12]. The purpose of the characterization model of a color
measurement instrument is to predict color measurement data from a given set of reference
data (training set). Essentially, the derived model according to polynomial fitting technique
describes the colorimetric relationship between a given sample set of reference data and
the corresponding measurements taken by an instrument. Consequently, the derived model
is applied to another set of measurements (test set) obtained by the same instrument. The
obtained new data set is a corrected version according to the used regression model, as
depicted in Figure 2. In this work, for each instrument a separate model has been derived
and consequently applied to the test set to correct the measurement data set. It is assumed
that by modeling the systematic errors of the measurement instruments the results will
improve the colorimetric performance and hence the inter-instrument and inter-model
agreement. In other words, the color difference between two corrected measurement data
sets will be reduced.
BCRA tiles Training set BCRA tiles

Reference data Measurements
Instrument 1
Regression model
Color wedge Corrected

Measurements Measurements
Test set
Figure 2. Schematic diagram of the regression method using a training set and a test set.
295
Paper G
A set of 14 British Ceramic Research Association (BCRA) Ceramic Color Standards

Series II (CCS II) ceramic gloss tiles including one Black and one White BCRA tile and
printed substrates were measured using six spectrophotometers, according to the
procedures outlined by ISO 13655 [16]. The instruments used are commercial industrial-
oriented spectrophotometers typical utilized for daily production control in prepress and
press applications. A spectrophotometer measures the ratio of reflected to incident light
(the reflectance) from a sample at many points across the visible spectrum [5]. Table 1
presents the instruments employed including the corresponding specifications. Some of the
instruments are typically from the same model and some of them represent different
models from different manufacturers. According to Nussbaum et al. [22] the instruments
used in this study show an acceptable performance in terms of repeatability and
reproducibility.
Table 1. Overview of the six instruments used in this work and the corresponding
specifications.
Measuring Apert- Manufact- Spectral Manufacturer’s

without ure Measuring urer’s range claimed
replacement geometry claimed and short term
mm precision interval repeatability
Master Mean ΔE*ab 0.02

instrument ΔE*ab 0.3 380nm (Standard shift
Max ΔE*ab to from 10
Yes
4.0 45°x:0° 0.8 on 12 730nm measurements at
Secondary
BCRA at 10nm 10 sec. interval on
E
tiles white)
Mean
ΔE*94 < 380nm
Secondary 1.0 to
No 2.0 45°x:0° ΔE*94 < 0.2
A on 12 780nm
BCRA at 10nm
tiles
ΔE*ab 0.02
Mean
380nm (Standard shift
ΔE*ab 0.3
Secondary Yes to from 10
4.5 45°x:0° on 12
D 780nm measurements at
BCRA
at 10nm 10 sec. interval on
tiles
white)
Secondary Mean ΔE*94 < 0.1
B ΔE*94 0.4 380nm (From 10
Max ΔE*94 to measurements at 3
Yes 1.0 on 12 780nm sec. interval on
Secondary 4.5 45°x:0°
BCRA at 10nm white)
C
tiles
296
Paper G
One particular measurement instrument has been defined as a reference and is called the
‘master instrument’. It is worth pointing out that the chosen reference instrument is not
meant to represent the best or ideal instrument but instead to be a state to which we
compare the measurements obtained by other instruments. The corrected measurements
obtained from the regression model were then used as the starting point with which the
corrected measurements from other instruments were compared to find the polynomial
which gives the least color difference. The other five instruments are called ‘secondary’s’.
Note that because one of the manufacturers is requesting not to publish their name the
devices were anonymized by identifying the instruments as ‘master instrument’,
‘secondary A’, ‘secondary B’ …’secondary E’.
2.1 Experimental procedure

To create the characterization model a training set is required. This should consist of a
reference data set and the corresponding measurements performed by a measurement
instrument. As seen above, ASTM [3] defines accuracy as the conformance of a series of
measurements to the accepted value for a given sample. In other words how closely an
instrument can conform to a certain reference. In this study the reference values have been
provided by the manufacturer of the BCRA tiles. All measurement instruments have
performed first a normal calibration procedure according to the manufacturer’s
recommendations before measuring the tiles 15 times in a sequence.
Note that not only the precision between the instruments has to be improved but also the
accuracy in terms of the appropriate colorimetric values. Therefore, the BCRA tile
reference values, which are traceable, have been used to establish the model. The idea of
using the 14 BCRA tiles for the training set is to make the appropriate adjustments in terms
of the accuracy. Eventually the derived model has been tested using color measurements of
46 patches of the UGRA/FOGRA Media Wedge CMYK [26] on three different types of
printed substrates (Figure 3).
The first paper substrate was a hard-copy digital proof print, printed according to the ISO
12647-7 graphic art standards for paper type 1 simulation by a commercial printing house.
The second paper substrate was paper type 1 printed by the same commercial printing
house aiming at the ISO 12647-2 graphic art standards. And the third paper substrate was
paper type 5 Altona Test Suite reference print [23]. The CIELAB target values of the
UGRA/FOGRA media Wedge CMYK are based on print conditions as stated in ISO
297
Paper G
12647-2 to and the appropriate characterization tables for different paper types are
provided by Fogra [19].
Figure 3. UGRA/FOGRA Media Wedge CMYK.
2.2 Polynomial regression

Essentially, polynomial regression is needed whenever you have two data sets, which are
related and you are aiming to predict one of them if you know the other, typically device
dependent colors from device in-dependent color and vice versa. Polynomial device
characterization technique with least squares fitting for different application has been
adequately explained and applied in a number of studies by Kang [18], Sharma [27], Hong
et al. [13] and Johnson [17]. However, a brief description is given to demonstrate this
technique applied to color measurement instruments.
The first data set, on which the regression model is based, is referred to as the training set
and the second data set, which was not involved in deriving the model is used for
evaluating it, and is referred to as the test set. As depicted in Figure 2 certain BCRA tiles
sample colors are selected and defined as the reference values and the measurement
instruments used were measuring the corresponding color specifications. For simplification
purposes let’s call the BCRA reference values ‘REF’ and the corresponding measurement
values from the instrument ‘INS’.
Assume that the reference target has N samples. For each color sample the corresponding
reference values R, E and F are represented by a 1 x 3 vector pi (i = 1…N) and their
corresponding I, N and S color measurement values obtained by the measurement
instrument are represented by a 1 x 3 vector xi (i = 1…N). Suppose that only I, N and S
values are used in p, the transformation between INS and REF is a simple linear transform.
However, the reason for using polynomials is that vector pi can be extended by adding
more terms such as I2, N2, S2 etc. which may improve the accuracy of the model in terms
of reducing the color differences over all the color samples [13]. While higher order
polynomials will give a perfect fit to the data of the training set, this may result in over
fitting as well as causing oscillations between the points, Runge's phenomenon [28].
Hence, in this work we applied only second order polynomial even for a 3 x 11 matrix.
298
Paper G
The polynomials applied and analyzed in this work have the following form:
1. pi = [I N S]
2. pi = [I N S 1]
3. pi = [I N S INS 1]
4. pi = [I N S IN IS NS]
5. pi = [I N S IN IS NS INS 1]
6. pi = [I N S IN IS NS I2 N2 S2]
7. pi = [I N S IN IS NS I2 N2 S2 INS 1]
Suppose R denotes an 3 x N matrix of vectors pi and X the predicted matrix of vector xi.
The mapping from INS to REF can be expressed by
X = M*R (1)
M is the unknown transformation matrix that determines the accuracy of the model, which
means minimizing the color differences over all color samples. The differences between Y’
and Y can also be expressed as the Sum of the Squares of the Differences (SSD):
(2)
where Y’= MX
(3)
Depending on the polynomial being solved the size of the matrix M in this work varies
from 3 x 3 up to 3 x 11. On the following 1st order sample it is shown how the model can
be derived.
Forward model: [R E F] = [I N S IN IS NS INS 1]*M
In this case M is an 8x3 transformation matrix that contains the model parameter
calculated from the training set by the equation:
M = (RT*R)-1*RT*X (4)
where RT denotes the transpose of R, and R-1 denotes the inverse. In this example R is an n
x 8 matrix which contains values of the I N S samples as well as corresponding IN, IS, NS,
INS and 1 values calculated from them for each sample. X is an n x 3 matrix which
contains the number of samples n used in the training set and the columns accommodate R,
E and F values of all the samples.
299
Paper G
The regression model is based on the training set containing 14 BCRA reference values
and the corresponding measurements obtained by the measurement instruments from the
BCRA tiles. The performance and accuracy of the characterization model has to be
evaluated using an independent data test set, which in this work is represented by
measurements of the 46 patches (UGRA/FOGRA Media Wedge) on different substrates.
The best results with the least color differences are obtained by experimentation.
Finally, the most appropriate transformation matrix M is the one that results in the least
color difference between the corrected measurements of the ‘master instrument’ and the
corrected measurements of the ‘secondary instruments’. Note that in this study CIELAB
values are directly used in the characterization and evaluation procedure because CIELAB
values have been reported from the spectral reflectance data initially measured by the
instruments. Furthermore, the ISO tolerances given in the standards are communicated in
CIELAB color space as well. Moreover and most important, Euclidian distance in
CIELAB color space is corresponding quite well to the perceptual color differences [12].
2.3 Data collections

All instruments used in this study measured spectral reflectance factor values from 380nm
to 730nm with 10nm intervals. Spectral measurements were converted to CIEXYZ
tristimulus values according to the CIE 1931 2° observer and the CIE Standard illuminant
D50 using the method proposed by ASTM 308, Table 1 [1]. Furthermore, to use a visually
meaningful color space CIELAB (D50 as the reference white) values were calculated
according to CIE 15 [10] specifications. Consequently, CIELAB data have been used for
the regression model and colorimetric difference ΔE*ab values were computed between the
master measurement instrument and the secondary instruments. Furthermore, the obtained
results will be compared with the ISO tolerances. Because the colorimetric production
control tolerances in the ISO standard 12647-2 and ISO standard 12647-7 are defined with
ΔE*ab only, no further color difference metrics are used in this work.
3. RESULTS AND DISCUSSIONS
As mentioned previously the aim is to find a method to reduce the color difference
between instruments measuring the same color patches. Furthermore the applied model
shall improve the colorimetric performance and inter-instrument and inter-model
agreement on three different types of substrates.
300
Paper G
Figure 4 shows the color difference results between the BCRA tiles reference values and
the ‘master instrument’. Moreover, the color difference results between BCRA tiles
reference values and the ‘secondary instrument A’ and between the ‘master instrument’
and the ‘secondary instrument A’. Notice, that the ‘master instrument’ and the ‘secondary
instrument A’ are not from the same instrument family (which in a practical application
very often can be the case). We see that the color difference between the BCRA tiles
reference values and the ‘master instrument’ has the highest values in the red (ΔE*ab 2.8
units), orange (ΔE*ab 4.2 units) and bright yellow (ΔE*ab 2.4 units) tiles. Comparing the
BCRA tiles reference values with the ‘secondary instrument A’ only the bright yellow tile
shows a rather high color difference value (ΔE*ab 2.4 units). On the other hand, comparing
the measurement results between the ‘master instrument’ and ‘the secondary instrument A’
the results on the red, orange and bright yellow tiles again show very large color
differences. Moreover, although the ‘master instrument’ and the ‘secondary instrument A’
show almost identical color difference compared to the BCRA tiles reference values on the
bright yellow tile (approximately ΔE*ab 2.4 units), the direct comparison shows the largest
color difference of ΔE*ab 4.5 units. This indicates that the accuracy of both measurement
instrument, ‘master instrument’ and the ‘secondary instrument A’ on the bright yellow tile
can be considered as very similar, However, the color difference between the ‘master
instrument’ and the BCRA tiles reference values and between the ‘secondary instrument
A’ and the BCRA tiles reference values points in different directions.
It is important to consider the inherent physical properties of the BCRA tiles. Fairchild and
Grum [11] stated that the BCRA tiles red, orange and yellow can exhibit appreciable
thermochromism due to sharp changes in their spectral reflectance curves. Based on this
finding Berns [6] argued against using the tiles red, orange and yellow unless the
temperature of the tiles at the time of calibration was known and this temperature was
maintained both at the location where the tiles would be used and during their
measurements. However, according to the results shown in Figure 4 there is no clear
evidence of thermochromism for the ‘secondary instrument A’ except for the yellow tile.
In contrast, the ‘master instrument’ demonstrates larger color differences due to possibly
generating significant heat in the measuring process. According to Fairchild and Grum
[11], it is important to make sure that the temperature of calibration standards remains
constant during their use. On the other hand, no significant color changes have been
observed with small temperature changes around room temperature. Furthermore, all
301
Paper G
measurements in this study have been conducted in the same location and the same room
temperature conditions.
Figure 4. Color difference between the 14 BCRA tiles and the ‘master instrument’, BCRA
tiles and the ‘secondary instrument A’ and between the ‘master instrument’ and the
‘secondary instrument A’.
The impact of such a systematic error due to the nature of different instrument properties is
illustrated in Table 4 which shows the color difference results between the measurements
performed by the ‘master instrument’ and the ‘secondary instrument A’. The
measurements were conducted by measuring the control media wedge on proofing
substrate. Additionally, the table indicates the defined ISO tolerances according to ISO
12647-7. It can be seen that the ‘master instrument’ gives values which qualify the proof as
approved. On the other hand, the ‘secondary instrument A’ gives values which are outside
the given tolerance (orange marked values) and therefore the proof might be not accepted.
Moreover, the maximum color difference is clearly seen in the yellow color, which exceed
the color difference tolerance of ΔE*ab 5 units. This is perhaps not unexpected due to the
large difference in terms of the accuracy performance between the two measurement
instruments on the bright yellow BCRA tile, seen previously. Nevertheless, the question
may arise, whether the proof is really generated very close to the standard values, or it is
just outside the colorimetric tolerance obtained by the two measurement instruments,
which varies in terms of accuracy conformance.
In a real practical application this situation can be considered as very inappropriate, where
due to systematic errors of the measurement instruments the one instrument results in
approved and the other one not. Therefore in a first attempt the measurement instrument
302
Paper G
accuracy has to be ensured by modeling the relationship between the BCRA tile reference
values and the actual measurements performed by the measurement instruments.
Table 2. Results from the regression using CIELAB values according to 14 BCRA tiles
references values and the corresponding values of the ‘master instrument’ and the
‘secondary instrument A’.
Trainings ‘Master instrument’ BCRA (reference) BCRA (reference)
data set 14 versus versus ‘Master versus ‘Secondary
BCRA tiles ‘Secondary instrument A’ instrument’ instrument A’
Average Max Average Max Average Max
ΔE*ab ΔE*ab ΔE*ab ΔE*ab ΔE*ab ΔE*ab
Real ΔE*ab 1.47 4.45 1.07 4.17 0.91 2.39
Matrices
3x3 0.44 1.20 0.43 1.02 0.29 0.63
3x4 0.45 1.10 0.44 1.01 0.28 0.64
3x5 0.43 1.07 0.34 0.84 0.21 0.41
3x6 0.29 0.49 0.18 0.25 0.16 0.25
3x8 0.15 0.30 0.12 0.25 0.09 0.20
3x9 0.11 0.39 0.10 0.21 0.06 0.21
3 x 11 0.05 0.12 0.06 0.18 0.03 0.10
Table 2 shows the results of the training set with 14 samples using polynomial regression
minimizing the difference between reference values and corresponding measurements. As
expected, the higher the degree of the polynomial the more reduction in the difference in
terms of average ΔE*ab and maximum ΔE*ab. The regression is a function of CIELAB
reducing differences in systematic errors to extremely low levels.
The derived training functions are used to correct the colorimetric values of the
UGRA/FOGRA Media Wedge measured by the ‘master instrument’ and the ‘secondary
instrument A’ on three different type of substrates. To determine the appropriate function
in terms of the least color difference between the corrected measurement data conducted
by the ‘master instrument’ and the ‘secondary instrument A’, all calculated training
functions have been applied to the test data of the 46 color patches of the UGRA/FOGRA
Media Wedge. Figure 5 shows the results (Mean and Max ΔE*ab) on proof substrate using
different correction matrices on a 3-D surface. The horizontal axis and the depth axis
303
Paper G
indicate the polynomials used for the correction of the ‘master instrument’ and the
‘secondary instrument A’. The vertical axis point out the corresponding corrected color
differences. It is important to note, that the uncorrected color difference between the
‘master instrument’ and the ‘secondary instrument A’ on proofing paper results in Mean
ΔE*ab 2.59 units and Max ΔE*ab 6.92 units respectively.

Figure 5. Results from the regression applying to the test set of the ‘master instrument’
and the ‘secondary instrument A’ on proofing substrate.
Further, it can be seen from Figure 5, that the systematic error of the ‘master instrument’
and the ‘secondary A’ can be corrected and the color differences reduced by applying
almost all functions to the uncorrected measurement data. The goal is to find a connection
between modeling the least color difference and the number of terms in the matrices. As
can be seen there are a number of functions, from simple 3 x 3 to 3 x 9 polynomials, which
reduce the color differences with more than half compare to the uncorrected data.
Although, as seen in Table 2, higher polynomial functions perform excellent in the training
set, testing the functions on the test set, the results from the test set indicate that using the
higher polynomials (e.g. 3 x 11) on the training set has resulted in over fitting. Therefore, it
is generally recommended to use the smallest number of the polynomial terms which
adequately fits the function while still smoothing out the noise [4]. Though, there is no
single function performing significant best the 3 x 4 or 3 x 5 polynomials can be
considered as most appropriate for both the ‘master instrument’ and the ‘secondary
instrument A’.
To find the relationship of the color differences in terms of different color attributes
between the ‘master instrument’ and the ‘secondary instrument A’ of the uncorrected and
304
Paper G
corrected data set, ΔE*ab versus lightness (L*), ΔE*ab versus chroma (C*ab), and ΔE*ab
versus hue-angle (h*ab) are plotted in Figure 6.
Figure 6. Color difference distribution between the ‘master instrument’ and the ‘secondary
instrument A’ of the uncorrected and corrected data set: plot of lightness vs. ΔE*ab; plot of
chroma vs. and ΔE*ab ; plot of hue angles vs. ΔE*ab.
The results of the corrected data set were obtained when a 3x5 matrix was used. Figure 6
illustrates the color difference distribution between the uncorrected and the corrected data
set, which gradually affect the color attributes lightness, chroma and hue angle. It can be
seen that larger color differences, especially in highly saturated colors, have been reduced
significant by the correction model. As expected, the largest color difference in the
uncorrected data set is in the yellow color (ΔE*ab 6.92 units). After the correction is
applied to the measurements of the ‘master instrument’ and the ‘secondary instrument A’
the color difference is reduced to ΔE*ab 2.24 units.
To further evaluate the performance of the proposed model the same procedure has been
applied for substrate paper type 1 and paper type 5. Moreover, the proposed technique has
been used and tested for all the ‘secondary instruments’. Table 3 presents the results from
the regression and the corresponding color difference between the ‘master instrument’ and
the ‘secondary instruments (A-E)’ on different types of paper substrates. In addition, the
color differences between the ‘master instrument’ and the ‘secondary instruments (A-E)’
of the uncorrected measurement data are presented. Except for ‘secondary instrument A’
the color differences of the uncorrected measurement data are smaller between the other
‘secondary instruments (B-E)’ and the ‘master instrument’. This is due to the different
instrument product family. However, in addition to the proofing substrate the model is
performing very well in terms of reducing the color differences on the two other substrates.
As discussed previously, there is no polynomial which performs significantly best. Very
small differences in the results among the functions can be observed. Similar performance
305
Paper G
patterns in terms of the size of the function can be found from the other ‘secondary
instruments (B-E)’. Note, that the polynomials for both the ‘master instrument’ and the
‘secondary instrument (A-E)’ have almost the same size for all three substrates. For the
proofing substrates, the applied correction has reduced the color difference both in terms of
mean and maximum for all the ‘secondary instruments (A-E)’. The same is true for the
results from the substrate paper type 5. For substrate paper type 1 and the ‘secondary
instrument B’, ‘secondary instrument D’ and ‘secondary instrument E’ the correction
model is not performing as good (slightly higher maximum values).
Table 3. Results from regression and the corresponding color difference between master
instrument and the secondary instruments.
Master ΔE*ab ΔE*ab
instrument Type Un-corrected Polynomial Corrected
versus Substrate Max Mean Master Secondary Max Mean
Proofing 6.92 2.59 3x4 3x5 2.67 1.17
Paper type 1 6.14 2.36 3x4 3x4 2.09 0.98
Secondary A Paper type 5 4.38 1.72 3x5 3x5 2.34 1.18
Proofing 1.21 0.47 3x5 3x5 0.70 0.34
Paper type 1 1.28 0.47 3x5 3x5 1.59 0.43
Secondary B Paper type 5 1.25 0.51 3x5 3x5 1.09 0.41
Proofing 4.67 2.48 3x4 3x6 4.49 2.26
Paper type 1 4.67 2.18 3x4 3x5 4.48 1.89
Secondary C Paper type 5 1.56 0.82 3x8 3x8 1.42 0.71
Proofing 1.41 0.48 3x5 3x5 1.07 0.41
Paper type 1 1.08 0.56 3x6 3x6 1.59 0.75
Secondary D Paper type 5 1.82 0.77 3x5 3x4 1.71 0.94
Proofing 0.62 0.33 3x4 3x3 0.61 0.25
Paper type 1 0.66 0.32 3x5 3x5 0.83 0.28
Secondary E Paper type 5 0.67 0.34 3x6 3x6 0.60 0.32
As indicated previously instruments can be divided into product families which are
instruments of the same model from the same manufacturer using equal specifications.
Hence, the least inter-instrument color differences can be expected within a product
family. In this work, the ‘secondary instrument E’ is the same model as the ‘master
instrument’. Regardless of the very small color differences of the uncorrected
306
Paper G
measurement data, applying the regression model has further minimized the color
differences. It is noticeable that, taking the inter-instrument agreement from Table 1 into
account which specifies the equipment accordance with mean ΔE*ab 0.3 units and
maximum ΔE*ab 0.8 units on 12 BCRA tiles ceramics the proposed method is performing
reasonable. On the other hand, the model is not able to handle the instrument’s
repeatability issues such as drift over time.
The ‘secondary instrument B’ and ‘secondary instrument C’ are considered as the same
instrument model too. However, the ‘secondary instrument C’ shows rather large color
differences, in particular the values given for proofing substrate and substrate paper type 1.
The reason is an UV cut filter attached to the instrument which causes the color differences
due to the concentration of optical brighteners to affect the CIE b* values in the
measurements. Such variations are not considered as systematic errors. Therefore, the
applied model is not performing as expected, in terms of reducing the color differences.
In the context of quality control using more than one measurement instrument in the
workflow, the proposed method can improve the inter-instrument and inter-model
agreement significant. Table 4 shows the color differences on proofing substrate according
to the uncorrected measurement data of the master instrument and the secondary
instruments A-E. Furthermore, the orange marked numbers demonstrate the values, which
are outside the tolerances defined by ISO 12647-7. According to the results presented in
Table 4 the measurements from the ‘secondary instrument A’ and the measurements from
the ‘secondary instrument C’ the proof would not be qualified as approved. On the other
hand, measurements conducted with the ‘master instrument’, the secondary instruments B,
D and E the proof would be qualified as approved. Again, the question may arise which of
the instrument gives the appropriate results? Note, that the instruments random errors
including repeatability performance has been tested in a previous study and concluded as
acceptable [22].
307
Paper G
Table 4. Color difference results on proofing substrate obtained by six instruments with
respect to CIELAB ΔE*ab tolerances according to ISO 12647-7 (Orange marked values
are outside the ISO tolerance).
Composed
Substrate Mean Max Primaries
grey
Un corrected
ΔE*ab ΔE*ab ΔE*ab ΔE*ab ΔH*ab ΔH*ab
measurements
3 3 6 5 2,5 1,5
C M Y K C M Y Average
Master 1.69 1.28 3.00 0.90 1.51 0.66 1.2 0.48 1.36 0.04 1.08
instrument
Secondary A 1.40 2.54 7.5 2.96 3.03 7.5 1.36 2.56 0.05 0.71 0.55
Secondary B 1.40 1.12 2.67 0.66 1.07 1.48 1.10 0.31 0.92 0.06 0.71
Secondary C 6.34 3.04 6.34 3.27 2.36 2.49 1.68 3.14 2.19 0.33 3.47
Secondary D 0.92 1.26 2.46 0.87 1.17 2.05 1.04 0.26 0.86 0.15 0.71
Secondary E 1.52 1.4 3.12 1.43 1.70 0.9 1.38 1.00 1.59 0.45 0.93
Table 5 presents the results of all instruments after applying the regression model to the
uncorrected data set. Although the ‘secondary instrument A’ results in values which now
qualify the proof as approved it is important to emphasize that it is not the intention of the
proposed method to get the values as close as possible to the ISO standard values but to
reduce the color difference between the instruments.
Furthermore, it can be seen that the ‘secondary instrument C’ with the UV cut filter still
results in values which qualifies the proof far from approved. The variations between the
‘secondary instrument C’ and the other instruments are large, especially the results
obtained on the substrate and the composed grey. As stated previously, this effect of
variation is considered as a systematic error and therefore the regression method is not
handling this.
308
Paper G
Table 5. Corrected measurement data set (based on 14 training samples) of seven

instruments with respect to CIELAB ΔE*ab tolerances according to ISO 12647-7.
14 training Composed
samples grey
Corrected ΔE*ab ΔE*ab ΔE*ab ΔE*ab ΔH*ab ΔH*ab
Measurement 3 3 6 5 2,5 1,5
data set C M Y K C M Y Average
Master 1.56. 1.55 3.22 1.96 1.81 2.87 1.25 0.57 0.98 0.19 0.78
(3x4)
Secondary A 0.92 1.83 4.42 2.12 1.85 4.56 1.17 2.00 0.97 0.13 0.41
(3x5)
Secondary B 1.40 1.33 3.39 0.90 0.94 2.62. 1.17 0.65 0.31 0.57 0.59
(3x5)
Secondary C 6.04 3.01 6.04 1.90 2.90 3.89 1.64 1.68 2.82 0.35 3.20
(3x6)
Secondary D 0.98 1.22 2.97 0.91 0.76 2.36 1.00 0.74 0.22 0.51 0.64
(3x5)
Secondary E 1.45 1.60 3.40 1.80 2.05 2.75 1.4 0.28 1.56 0.27 0.68
(3x3)
So far the training set for building the model was limited to 14 samples (14 BCRA tiles).
To test if the model will improve the performance in terms of reducing the color difference
between the ‘master instrument’ and the ‘secondary instrument (A-E)’ the sample number
of the training set has been increased with 24 patches from the ColorChecker, which is a
color rendition chart including traceable reference values [20]. Consequently, the
regression method has been applied again for all the measurement on all three substrates.
Although there is no significant improvement in terms of reducing the mean color
difference, the maximum ΔE*ab could be reduced substantially in all instrument
combinations and all three substrates. Moreover, also functions with second order
polynomial (such as 3x11) give reasonable results reducing the color difference
significantly, in particular the maximum color difference. This indicates clearly that the
model with 38 sample points is more robust.
309
Paper G
Table 6. Corrected measurement data set (based on 38 training samples) of seven

instruments with respect to CIELAB ΔE*ab tolerances according to ISO 12647-7.
38 training Composed
samples grey
Corrected ΔE*ab ΔE*ab ΔE*ab ΔE*ab ΔH*ab ΔH*ab
Measurement 3 3 6 5 2,5 1,5
data set C M Y K C M Y Average
Master 0.60 1.24 3.05 2.16 1.19 2.38 1.37 1.42 0.09 0.92 0.56
(3x4)
Secondary A 1.00 2.02 4.29 3.00 1.94 4.02 1.68 2.69 1.02 0.30 0.39
(3x3)
Secondary B 0.46 1.30 3.16 2.01 1.62 2.75 1.35 1.42 0.03 1.12 0.39
(3x4)
Secondary C 4.89 2.39 4.89 1.42 1.90 3.39 1.78 1.36 1.77 1.06 2.49
(3x5)
Secondary D 0.80 1.35 3.57 2.48 1.06 2.87 0.86 1.81 0.24 1.36 0.35
(3x3)
Secondary E 0.52 1.38 3.62 1.08 0.84 3.17 1.28 0.89 0.41 1.51 0.31
(3x6)
Table 6 shows the results using the training set with 38 samples. It can bee seen that,
except for the ‘secondary instrument C’ on composed grey tolerance, all corrected
measurement values obtained will qualify the proof as approved. In other words,
increasing the number of training samples in the model and applying the function to the
test set (in our case to the measurements of the proofing substrate) will correct the
measurements and reduce further the maximum color differences. It is important to note
that by increasing the number in the training set, the model behaves more robustly in terms
of using higher polynomials in correcting the test data set.
310
Paper G
Figure 7. Range of variations with respect to uncorrected and corrected measurement data
sets from all six instruments on proofing substrates.
Another way to describe the range of variation between the uncorrected and the corrected
measurements is standard deviation σ. As mentioned previously the aim of the presented
work is to reduce the variations related to color differences between measurement
instruments in measuring the same color patches. Figure 7 illustrates the range of
variations between the uncorrected and corrected measurement data sets (including 14
training samples and 38 training samples) on proofing substrates in terms of the standard
deviation (according to the results given in Table 4, Table 5 and Table 6). It can be seen
that the applied regression model is reducing the range of color difference variations
among the six instruments compared to the uncorrected data sets. In particular the
maximum values could be reduced significantly and therefore the mean results have been
affected too. Looking at color difference results for the substrate it can be seen that the
range of variations between the uncorrected data set and the corrected data set remains
almost the same. This is again due to the ‘secondary instrument C’ with the UV cut filter,
as explained above. Except for cyan and black, the correction model based on the training
set with 38 samples is performing better compared to the training set with 14 samples.
To further verify the proposed method, two other commercial measurement instruments
from different manufactures representing different models have been used. The first is
considered as a spectrophotometer and the second as a spectrocolorimeter. ASTM E1347
defines a spectrocolorimeter as a spectrometer that provides colorimetric data, but not the
underlying spectral data [2]. For the training set, 38 samples (14 BCRA tile and 24
311
Paper G
ColorChecker) including the reference values and the corresponding measurements have
been used to derive the model. The test data set has been extended from 46 patches
(UGRA/FOGRA Media Wedge CMYK, version 2.2) to the new 72 patches
(UGRA/FOGRA Media Wedge CMYK, version 3.0) and the measurements have been
conducted on proofing substrates. Table 7 presents the results correcting the measurements
of both instruments according to the training set for each device. As can be seen from the
results, the color difference between the two instruments has been reduced by
approximately 30% in both the mean and maximum respectively. The least difference has
been obtained by using a 3x5 polynomial for both instruments and indicating that the
applied model performs reasonable.
Table 7. Color difference results from the uncorrected and corrected measurements
between two different instruments on proofing substrate.
ΔE*ab ΔE*ab
Spectrophotometer Type
Un-corrected Polynomial Corrected
versus Substrate
Max Mean A B Max Mean
Spectrocolorimeter
Proofing 3.64 1.73 3x5 3x5 2.32 1.21
It has to be mentioned that only one single measurement with each instrument on the
training samples (14 BCRA tiles and 24 ColorChecker) has been conducted. Presumably,
averaging multiple measurements per sample will reduce the noise, increase the
performance of the model, and further reduce the color differences between the two
corrected test data sets from each instrument.
4. CONCLUSIONS
It is known that the accuracy and inter-instrument and inter-model agreement of

measurement instruments are limited. In this work we have described a method to correct
the instrument’s systematic errors by applying a regression technique directly onto the
measurement output values in the CIELAB color space to improve the colorimetric
performance and hence the inter-instrument and inter-model agreement.
The study compares different terms of polynomials derived using least-squares regression
to determine the appropriate correction for six different measurement instrument’s
measured on three different types of substrates. One of the measurement instrument used
312
Paper G
has been defined as the reference instrument. Reference data from 14 BCRA tiles and the
corresponding obtained measurements from each instrument has been used to derive a
model. The model has been applied to a test set containing 46 measurements from the
UGRA/FOGRA Media Wedge on three different substrates. To determine the most
appropriate polynomial color differences have been calculated between the corrected
measurements of the ‘master instrument’ the corrected measurement of the ‘secondary
instruments’. We conclude that first order polynomials (more precise 3x5 polynomial) in
most cases produce the best results in terms of reducing the color differences between the
instruments on different substrates.
Although there is no significant difference in the performance of the model on the three
different types of substrates, the proofing substrate results in the least color differences.
Moreover, with instruments from different product families the inter-model agreement can
be significantly improved by applying the characterization model, reducing the color
differences between the measurement instruments by more than 50%. Increasing the size
of the training set from 14 to 38 samples is slightly reducing the maximum color
differences, but much more important, the model’s behavior is more robust in terms of
different applied polynomials. To justify whether thermochromism affected the model, the
BCRA tiles red, orange and yellow could be left out in the training sample.
As seen, the proposed regression method works remarkably well with a range of
instruments used on the three types of substrates. However, for future work, the proposed
method could by further investigated using different paper substrates (e.g. glossy paper,
newspaper) and material (e.g. plastic, textile, aluminum, glass). To improve the
performance of the model further extension of the sample number including different types
of sample surfaces could be considered to derive the model. Furthermore, the method can
be extended and tested on emission measurements using different models of
spectrophotometers, spectrocolorimeters and colorimeters. Perhaps, a combination of
Bern’s [6] proposed method correcting various systematic errors in the spectral domain
and the presented technique adjusting the output CIELAB data set may further improve the
colorimetric performance and the inter-instrument and inter-model agreement. Finally,
considering a relevant application, the proposed model could be implemented into a
measurement software system where the correction model is directly applied to the
obtained measured values from the instruments.
313
Paper G
REFERENCES
the CIE-System,American Society for Testing and Materials, West Conshohocken,
PA, 2008
[2] ASTM E1347-06, Standard Test Method for Color and Color-Difference
Measurement by Tristimulus Colorimetry,American Society for the Testing of
Materials, West Conshohocken, PA, 2006
of Color Measuring Instruments,American Society for the Testing of Materials,
West Conshohocken, PA, 2008
[4] R. Bala, "Device characterization," in Digital color imaging handbook. vol. CRC
Press LLC, S. Gaurav. 2003.
York, 2000.
errors," Color Research & Application, vol. 13, pp. 243-256, 1988.
Applied Optics, vol. 8, pp. 775-783, 1969.
[9] J. Briggs, D. Forrest, M. Tse, and I. QEA, "Reliability Issues for Color
Measurement in Quality Control Applications," in IS&Ts NIP 1998, 14 pp. 595-
602.
[10] CIE15, Colorimetry,CIE Central Bureau, Vienna, 2004
Optics, vol. 24, pp. 3432-3, 1985.
[12] J. Hardeberg, Acquisition and Reproduction of Color Images, Colorimetric and
Multispectral Approaches Dissertation.com, 2001. ISBN 1-58112-135-0
[13] G. Hong, M. Luo, and P. Rhodes, "A study of digital camera colorimetric
characterization based on polynomial modeling," Color Research & Application,
vol. 26, pp. 76-84, 2001.
314
Paper G
colour separations, proof and production prints – Part 2: Offset printing
processes,ISO, 2004
color separations, proof and production prints – Part 7: Proofing processes
working directly from digital data,ISO, 2007
[16] ISO13655, Graphic technology – Spectral measurement and colorimetric
computation for graphic arts images ISO, 1996
[17] T. Johnson, "Methods for characterizing colour scanners and digital cameras,"
Displays, vol. 16, pp. 183-191, 1996.
[18] H. Kang, Color technology for electronic imaging devices: SPIE-International
Society for Optical Engineering, 1997.
[19] A. Kraushaar. Characterization data for standardized printing conditions 2008.
Available: http://www.fogra.de/
[20] C. McCamy, H. Marcus, and J. Davidson, "A color-rendition chart," J. App.
Photog. Eng ; now called “The Society for Imaging Science and Technology
(IS&T)”), vol. 2, pp. 95-99, 1976.
[21] K.-G. Nickel. (2010, November 10th ). Process Standard Offset, German Printing
and Media Industries Federation (bvdm) and Graphic Technology Research
Association. Available: http://www.pso-insider.de/index.php?id=48&L=1
[22] P. Nussbaum, A. Sole, and J. Y. Hardeberg, "Consequences of using a number of
different color measurement instruments in a color managed printing workflow," in
TAGA, 2009.
[23] Print & Media Forum AG, Altona Test Suite Anwendungspaket,Bundesverband
Druck und Medien, 2004
Barcelona, Spain 2008, pp. 77-79.
Application, vol. 19, pp. 322-331, 2007.
[26] U. Schmitt and F. Dolezalek, Ugra/FOGRA Media Wedge CMYK V 2.0,Munich,
2004
315
Paper G
[27] G. Sharma, Digital color imaging handbook: CRC, Press, New York, 2003.
[28] Wikipedia. (2010, retrieved 12. November 2010). Runge's phenomenon. Available:
http://en.wikipedia.org/wiki/Runge%27s_phenomenon
vol. 32, pp. 166-175, 2007.
316
Paper G
317

PHD Thesis-Peter

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PHD Thesis-Peter

Uploaded by

Copyright:

Available Formats

Colour Measurement and Print Quality Assessment

in a Colour Managed Printing Workflow

The Norwegian Color Research Laboratory

Submitted to the Faculty of Mathematics and Natural Science at the

Series of dissertations submitted to the

All rights reserved. No part of this publication may be

Cover: Inger Sandved Anfinsen.

Produced in co-operation with Unipub.

List of Included Papers

• P. Nussbaum, A. Sole, and J. Y. Hardeberg

• P. Nussbaum and J. Y. Hardeberg

• S. Roch, J. Y. Hardeberg, and P. Nussbaum

• J. Y. Hardeberg, P. Nussbaum, S. Roch, and O. Panak

The Publications ......................................................................................... IX

1.1 Motivation ................................................................................................3

2.1 Colour Fundamentals..............................................................................7

2.2 Colour Measurement.............................................................................27

2.4 Print Quality and Print Assessment.................................................... 55

3 Summary of included Papers ............................................................... 75

3.1 Paper A: Print Quality Evaluation for Governmental Purchase

3.2 Paper B: Print Quality Evaluation and Applied Colour Management in

3.3 Paper C: Implementing ISO 12646 Standards for soft Proofing in

3.4 Paper D: Print Quality Evaluation and Applied Colour Management

3.5 Paper E: A Comparison of Densitometric and Planimetric

3.6 Paper F: Analysis of Colour Measurement Uncertainty in a Colour

3.7 Paper G: Regression based Characterization of Colour

4.1 Discussion of Papers in Context .........................................................105

PART II INCLUDED PAPERS

1.2 Aim of the Study

important part of the study is the investigation of measurement devices in terms of

1.3 Research Methodology

1.4 Outline of Thesis

2.1 Colour Fundamentals

2.1.1 Human Visual System

Figure 1: Schematic representation of the principles of viewing colour.

2.1.3 CIE Standard Observer

Spectral tristimulus values

The tristimulus values X, Y, and Z for a given object, which is illuminated by a

Light source Object CIE Standard Observer Tristimulus values

X = k ∑ x ( λ )R( λ )S( λ )Δλ

2.1.4 CIE Standard Illuminants

2.1.6 CIE 1931 XYZ Colour Space

Although the CIE x, y chromaticity diagram is often used in colour management,

2.1.7 1976 Uniform Colour Spaces

Figure 6: CIE standardized uniform CIE u’, v’ chromaticity diagram.

-a* = Greenish +a* = Reddish

-b* = Blueish L* = 0 = Black

f(I)= (I)1/3, if I > 0.008856,

2.1.10 CIELAB Colour Difference

Figure 9: Schematic illustration expressing colour differences (after Morovic’s Figure on

ΔL* = L*sample - L* target

Δa* = a*sample - a* target

Δb* = b*sample - b* target

ΔE*ab = [(ΔL*)2 + (ΔC*)2 + (ΔH*)2]1/2, (9)

ΔH*ab = [(ΔE*)2 - (ΔL*)2 - (ΔC*)2]1/2, (10)

As mentioned previously, the perceived magnitude of the colour difference does

⎡⎛ ΔLab * ⎞ 2 ⎛ ΔCab * ⎞ 2 ⎛ ΔHab * ⎞ 2 ⎤1/ 2

2.2 Colour Measurement

2.2.1 Measurement Instruments and their Application

Standard Observer response

CIE geometries for illumination and viewing for reflection

• Diffuse: eight-degree geometry, specular component included (di:8°)

• Diffuse: eight-degree geometry, specular component excluded (de:8°)

ΔL* = Lsample - L target

Δa* = asample - a target

Δb* = bsample - b target

ΔEab = [(ΔL)2 + (ΔC)2 + (ΔH)2]1/2, (9)

ΔHab = [(ΔE)2 - (ΔL)2 - (ΔC)2]1/2, (10)