You are on page 1of 8

Comparing color image quality of four digital presses

Jon Y. Hardeberg and Sven E. Skarsb Gjvik University College, Gjvik, Norway,
Color image quality is a very important deciding factor for buyers of color imaging devices. It is therefore of utmost importance for manufacturers of imaging equipment such as digital presses, to pay special attention to this factor. We have carried out a study in which we have evaluated the color image quality of four different commercially available digital color presses, one of which is using a magnetographic technique for image formation; the other three use a more conventional electrophotographic technique. To evaluate color image quality we use a combination of several different techniques, based on colorimetric and spatial measurements, visual expert evaluations, and panel tests. We have found that there are signicant quality differences between the devices under test. Note, however, that we have tested specic total systems, including printer, RIP, parameters, ICC proles, paper, etc., and also that there are many important factors that are not in the scope of this test, such as price, printing speed, long term stability/expected lifespan, etc. Nevertheless, we believe that the results of this study is of signicance to both manufacturers and customers of digital press equipment. requirements, we believe that knowledge of the color image quality that different equipment can provide is of great importance to customers. In our study we have evaluated the color image quality of four different commercially available digital color presses, namely Oc CPS700, Xerox DC2060, Nexpress e 2100, and Canon CLC5000. Note that because of the obvious interest this will have to customers and others, we have decided to publish the names of the devices under test, similarly to what has been done in other benchmarking studies (den Engelsman, 2002, Lindberg et al., 2001, Bolanca et al., 2001). A popularized version of the results has been published in a Swedish graphic arts magazine (Skarsb and Hardeberg, 2002), and presented at two Nordic conferences. However, we also believe that not only the results, but also the methodology is of interest to the engineers and scientists working with printing technology. Electrophotography is the most widespread nonimpact-printing technology that exists. Three of the presses in the test (Xerox, Canon, and Nexpress) use variants of the electrophotographic principle with electrostatic powder toner, based on an invention of Chester Carlson from 1939 (Kipphan, 2001). The fourth press, Oc CPS700, uses a different princie ple that may be classied under the category magnetography even if Oc does not use that designation and calls e the method Direct Imaging Printing Technology. The imaging carriers are cylinders tted with individually controllable ring electrodes, protected by a dielectric coating. A magnetic, single component toner is fed to the imaging cylinders by a magnetic roller. The imaging is achieved by charging the ring electrodes with image-dependent voltage pulses. An imaging magnetic roller achieves the imaging by removing magnetic toner from the non-printing areas of the imaging cylinder. For multicolor printing with Oc CPS700, instead of e only using the usual four colors, it uses seven colors, CMYK plus the three complementary colors red, green and blue. Seven imaging cylinders are therefore positioned in a satellite conguration around one common intermediate cylinder, which transfers the toner to the paper. The multicolor printing is not achieved by overprinting of four colors. Instead, pixels of the seven colors are positioned alongside each other, and the press is thus relying

1. Introduction
After several years of market hesitation, digital presses have now become common. In todays printing market, where exibility, variable content, shorter lead times, and on demand publishing, are being demanded, digital presses represent an attractive supplement to conventional offset presses. It is important for customers of digital press equipment to be able to take into account the quality of the equipment, typically to be able to make a trade-off between price and quality. The total quality of a device is, however, a very complex entity, involving technical aspects such as expected lifespan, printing speed, accepted media, as well as customer relation aspects such as service agreements. Also customers who do not intend to own digital press equipment, but rather are looking for instance to have a publication printed by a print bureau, need to consider the quality of service they can get with different providers. While different customers have different

on additive color mixture. This may be the advantage as well as the disadvantage of the Oc direct imaging printe ing technology. Just one thin layer of toner can be xed with less heat supply than four layers. The mono-layer also prevents the brutal relief building typical for photoelectric color prints. On the other hand, pixels of different colors placed alongside each other tend to make smooth tones look grainy. In the remainder of this paper we rst give an overview of the eld of color image quality, in Section 2. We then proceed in Section 3 to a description of the experimental setup and results of our study in which the color image quality of four digital presses were evaluated. The results are further discussed in Section 4, along with a higher level discussion of factors surrounding this study. Finally, we round off by a conclusion and discussion of further work in the area of color image quality.

Tradeoff analysis of speed and implementation cost versus color image quality in image processing algorithm development. Benchmarking of imaging systems and algorithms to other vendors products. Documentation of color image quality improvements resulting from efforts spent on optimization of technology parameters. For customers, it would obviously be advantageous to have access to reliable and objective information about the image quality that devices can provide when considering several alternatives for purchasing.

3. Experimental Setup and Results

To carry out our color image quality evaluation, we rst designed two test targets. The paper size used was A3, the resolution 600dpi, and the le format PDF. The target shown in Figure 1 was specied using CMYK color space, while the one shown in Figure 2 uses the sRGB color space (Anderson et al., 1996, IEC 61966-2.1, 1999). The targets contain several graphical and pictorial elements which were used for our quality evaluation.

2. Color Image Quality

In recent years, the concept of image quality has received quite much attention within the imaging science and technology community. The subject has for instance been extensively discussed at the PICS conferences (see e.g. Stokes, 1998, Rasmussen et al., 1998, Yendrikhovskij, 1999, T pfer and Cookingham, 2000, Kane et al., 2000, o Jung et al., 2001, Engeldrum, 2001, Hardeberg, 2002), and at more graphic arts oriented conferences (see e.g. Lindberg et al., 2001, Bolanca et al., 2001, Norberg et al., 2002, Edinger, 2002). But still, image quality often receives a rather stepmotherly treatment in the industry probably because of its somewhat awkward position between subjectivity and objectivity (Yendrikhovskij, 1999). The concept of quality, typically dened in dictionaries as degree of excellence is inherently a subjective entity. An engineer and scientist, however, generally prefers to deal with objective quantities, backed by scientic evidence. On the web site of a major French consumer electronics retailer, a formula for the image quality of a printer was given approximately as follows: Image quality = Resolution x Color Depth. This is an example of another common misconception regarding image quality its oversimplication. There are indeed many factors that contribute to the quality of an image, such as spatial resolution, color depth, the nesses (sharpness, naturalness, colorfulness, etc.), and visual artifacts (banding, streaking, grain, blocking, mottle, moir , etc.). There exist an ongoing effort to stane dardize the denitions of these and other image quality factors, as well as their assessment methodology, see for example a recent paper by Grice and Allebach (1999). It is out of scoupe of this paper to give an extensive description of these quality factors. We will, however, present and discuss the factors included in our analysis, in Section 3. Potential uses of quantiable data on color image quality for manufacturers include the following:

Figure 1: The CMYK test target designed for our study.

The actual test printing was carried out in the manufacturers ofces in Norway, except for that of Nexpress, which was carried out at Heidelberg in Great Britain, since there were yet no such presses installed in Scandinavia. The presses were operated by the manufacturers own personnel, in the presence of the authors. The CMYK target was printed without color management, that is in particular, not with the intention of proofing or simulating another press. The sRGB target, however, was printed using ICC-based color management with four different rendering intents perceptual, saturation, media-relative colorimetric, and ICC-absolute colorimetric (ICC.1:2001-12, 2001). For each of these, 20 copies

representing the gamuts in CIELAB color space, using the convex hull method, as implemented in the ICC3D tool (Farup and Hardeberg, 2002, Hardeberg and Farup, 2002). Projections of the gamuts onto the ab-plane is shown in Figure 3.

Figure 2: The sRGB test target designed for our study.

were printed. The manufacturer then chose the rendering intent that they preferred, and 600 more copies were printed with this intent. After this, the press was turned off and restarted after at least 5 minutes, whereby a new set of 20 copies was printed. The manufacturers were allowed to choose printing parameters such as raster frequency, RIP software, and paper, with the goal of achieving the best color image quality. The quality analyzes were done at the Color Lab at Gjvik University College, and included quantitative analyzes based on measurements, psychophysical experiments, as well as expert evaluations. The quantitative analyzes included colorimetric measurements to determine color gamut (Section 3.1), colorimetric reproduction (Section 3.2) and stability (Section 3.3). The psychophysical evaluations (Section 3.4) were done to determine color pleasantness, total image quality, smoothness, and detail rendition. In the expert evaluation we examined the halftoning, text readability, and alignment. 3.1. Color Gamut The color gamut of a digital press (or any other imaging device) is the sum of all colors it can reproduce. Colors that are outside of this gamut cannot be reproduced. The color gamut, and in particular its size, is thus a quality factor. A press with a larger gamut than another is typically able to reproduce more saturated colors, which can be appreciated for many applications. The color gamut of a digital press depends on many factors such as toner/colorant, substrate, and halftoning algorithm. The color gamuts were quantied based on spectrophotometric measurements of the TC3.5 CMYK target (Figure 1. For the case of the Oc press, however, we used the e union of this data and data from the RGB target, since some colors were found to be attainable only in RGB mode. From these measurements, we created solid 3D objects

(a) Oc CPS700 e

(b) Xerox DC2060

(c) Canon CLC5000

(d) NexPress 2100

Figure 3: The color gamut of the four devices, projected on the ab-plane.

As a supplement to the visual appreciation of the gamut shapes and sizes, we also calculated its volume (CIE TC805, 2001), see Table 1. We see that the Xerox press has a larger gamut than the other electrophotographic presses, and that Oc CPS700 has a signicantly smaller gamut. e This is probably due to unwanted mis-coloring of the toner from the iron oxide particles, and it is expected that Oc is e working to improve this.
Table 1: Size of the color gamuts, quantied as the volume of the convex hull of the gamut in CIELAB color space.

Device Canon CLC5000 Xerox DC2060 Nexpress 2100 Oc CPS700 e

CIELAB volume 437000 480000 345000 269000

Relative volume 91% 100% 72% 56%

As references for gamut size calculations, it can be mentioned that the gamut volume of the sRGB color space is found to be approximately 821000, and all the devices under test thus have signicantly smaller gamut volume than sRGB. Even so, there are parts of color space where the gamuts of all the presses exceed that of sRGB. Klaman (2001) evaluated the gamut volume of different presses by using a dodecahedron with the vertices dened by the primary and secondary colors, black and white. For different types of presses, she found gamut volumes ranging from approximately 100000 to 350000, with electrophotography situated in the range of 160000 to 210000. Because of the methodological difference in how to quantify the gamut volume, these numbers are not really relevant to our study, however.

difference, are not met, the device does not receive Microsofts certication the designed for Windows logo. The color patches of the 5x5x5 target (see Figure 4) on the prints were then measured with a GretagMacbeth SpectroScan/Spectrolino spectrophotometer, and the values were analyzed using a Microsoft Excel spreadsheet developed on the basis of Microsofts recommendations (Microsoft Corporation, 2001). The ideal reproduction is dened according using colorimetry that is relative both to the paper white and to the media black. The deviation between the printed colors and ideal colors were specied as E94 .

3.2. Colorimetric reproduction In some cases accurate reproduction of colors is important for the customer, for example when printing color samples or when digital prints shall be used as proof prints and the aim is to match offset prints. To investigate how accurate the presses can reproduce dened colors, 125 color patches dened in sRGB were printed on all the four presses using ICC-based color management and relative colorimetric intent. One means of evaluating color quality is thus through the measurement of color differences between the actual reproduction and a preferred color reproduction. The preferred color reproduction typically relates to an original document when evaluating color copy, the colors as they appear on the monitor for a typical WYSIWYG evaluation, or specic colors of which the color reproduction is particularly important for a given reproduction. The color differences are typically measured in terms of Eab the Euclidean distance between two colors in the CIELAB color space (CIE 15.2, 1986, Wyszecki and Stiles, 1982). Since natural images rarely contain sufciently uniform areas to allow consistent color measurements, a color target with several uniformly colored patches should be used. Simple statistical measures such as maximum and average color differences are then typically used as an indication of the color quality. Note that newer formulae for color difference are slowly replac ing the Eab , for instance E94 (McDonald and Smith, 1995). A very helpful resource for such color quality evaluation is the Microsoft Windows Color Quality Test Kit (Microsoft Corporation, 2001). This freely available kit contains descriptive documents, test targets and images, as well as tools for the calculation of color differences, for several different color imaging devices. Typically the goal is for the devices to communicate images using the sRGB color space (Anderson et al., 1996, IEC 61966-2.1, 1999). If certain criteria, in particular in terms of average color
Figure 4: The 5x5x5 RGB target used for evaluation of colorimetric reproduction accuracy. The colors are divided into two groups, the in-gamut colors, which are expected to be within the gamut of the digital press, and the gamut surface colors which have a large probability of being out of gamut.

To make an attempt to separate between the unavoidable colorimetric errors due to out-of gamut colors, and the errors on colors that could have been correct, we divided the target into two regions, as shown in Figure 4, the safe in-gamut colors, and those that lie on the surface of the sRGB space, and that are prone to being out of gamut of the press. The results are presented in Table 2. For in-gamut colors Oc obtained the best result, while Xerox got the highe est score for all colors seen together.
Table 2: Colorimetric accuracy, measured as average E94 color difference between printed color and an ideal sRGB reproduction dened according to Microsofts color quality specications (Microsoft Corporation, 2001). The media-relative rendering intent was used.

Device Canon CLC5000 Xerox DC2060 Nexpress 2100 Oc CPS700 e

In-gamut 10.3 7.5 9.2 6.5

Gamut surface 14.5 10.9 12.7 13.1

Total 13.6 10.2 11.9 11.7

3.3. Stability To get an indication of the short-term stability and repeatability of the devices, we measured two color patches before and after the device had been restarted. We report in Table 3 the Eab color difference between the averages of

10 measurements of the colors before and after restart. We see that the Xerox DC2060 and the Oc CPS700 have the e highest stability.
Table 3: Stability of the devices quantied as Eab color difference between averaged measurements of two patches printed before and after restart.

0,0 Excellent 0,5 1,0 Very good 1,5 2,0

IQ Score

Good 2,5 3,0 Fair 3,5 4,0

Device Canon CLC5000 Xerox DC2060 Nexpress 2100 Oc CPS700 e

Patch 1 1.24 0.56 1.63 0.44

Patch 2 0.61 0.39 1.47 0.58

Neutral 4,5 5,0 5,5 6,0

Color Pleasantness Total Image Quality Smoothness Detail rendition

Poor Very poor

Oce CPS700

Canon CLC5000

Xerox DC2060

NexPress 2100

Figure 5: Results of the psychophysical evaluation

3.4. Psychophysical evaluation To evaluate the visual quality that the presses can achieve, we carried out a psychophysical experiment, or in more common terms, a panel test. Seven observers were asked to evaluate three different images (Figure 2) according to four different quality criteria: color pleasantness, total image quality, smoothness, and detail rendition. For each quality criterion, the observers gave ratings on a scale from excellent (0) to very poor (6). The images were of course anonymized, presented in a pseudorandom sequence, and viewed under standard D50 illumination. The test showed us that typically, the images printed with saturation and perceptual intents were judged to be best, and that there was a correlation between total image quality and color pleasantness. The results of the experiment are summarized in Figure 5. The averaged result over the three images, and the best rendering intent, is used to represent a device. We see that for the three electrophotographical presses, the differences are small, perhaps with the exeption of smoothness, for which the Nexpress device came out signicantly better than the others. That the magnetographic press came out last can probably be attributed to two factors smaller color gamut and graininess because of the additive color mixture. 3.5. Halftoning To investigate the halftoning methods, we captured microscopic images on different locations on the target, see Figure 6. For comparison, we also captured a conventional offset print, with a raster frequency of 69 lpcm (150 lpi), as shown in Figure 7.
Figure 6: Microscopic enlargements of the raster structure of the four presses.

Not surprisingly, we see that Nexpress uses a raster structure that is very similar to that of offset, both with regards to dot shape and raster angles (15, 75, 0, and 45 degrees). Xerox also resembles offset, with Postscript angles (0, 18.4, 45, and 71.6 degrees). Canon is using essentially a line raster with angles of 105, 75, 50, and 90 degrees. Oc uses raster structures which appear to be AM/FM hye brids not surprisingly given its additive color mixture. 3.6. Summary of the results All the four presses have support for color management, and allows the user to choose between the four rendering intents dened by the ICC, although they are sometimes given different names in the user interface. Canon CLC5000 was judged to produce the most pleasing colors, closely followed by the Nexpress 2100. Concerning smoothness, the Nexpress device was a relatively clear winner, and it was also best on detail rendition, that is, what we might call visual resolution.

Figure 7: Microscopic enlargements of the raster structure of an offset press.

Xerox DC2060 has the largest color gamut, while Oc e CPS700 has the smallest, despite its seven colorants. Oc was best with regards to colorimetric accuracy of e in-gamut colors, while Xerox came out best when all the colors of the sRGB target were taken into consideration, probably because of its large color gamut. Xerox and Oc were the most stable devices, as observed by their low e color difference between the same patches printed before and after a restart of the device. The examination of the raster structures showed that Nexpress uses a raster which is nearly identical to traditional offset raster. Xeroxs raster is also similar to offset, while Canon, and to an even greater extent, Oc , have choe sen different approaches to halftoning. All the presses had good alignment between the different process colors, the readability of small positive and negative text was good, and there were no signicant errors in how the PDF le was printed. All in all, if we had to elect a best in test it would be the Nexpress device, but the distance down to the Canon and Xerox devices is small and hardly relevant. They are all capable of printing beautiful color pictures. The main problems with the Oc CPS700 is that it has e a much smaller color gamut than the electrostatic presses, and that the special additive color mixture, with pixels of different color side by side, gives a more grainy appearance than its competitors. This is most striking when the prints are compared visually to the other technologies, and much less problematic when looking at a print in isolation. The strength of the Oc device is that it only lays down a e thin layer of toner on the paper, and it is xated at low temperature with little strain on the substrate. It might be argued that our test was designed in such a way that it to a great extent revealed the weaknesses of this press, but that it did not appreciate its strong sides.

4. Discussion
Although the results of our evaluation presented in the previous section are indeed interesting, and the methods are valid in themselves, it is appropriate to proceed to a critical discussion. Concerning the choice of paper, it is obvious that for the sake of comparison, it would have been better if the dif-

ferent presses had used the same paper. The reason why we did not want to dictate the use of a certain paper, was that we would avoid forcing the manufacturer to use a paper that might not be well suited for their printing process, and thus that would have given sub-optimal results. We suggest that a better procedure would be to print on two paper stocks, one preferred by the manufacturer, and one common to all. With regards to the examination of the raster structure, we argue that even if it is interesting to study closeup views, comparing to offset raster, this should not be viewed as a quality criterion. It is much more relevant to look at how the images look at normal viewing distances. Are they smooth, with good color, good detail rendition, and without visible moir ? e It has previously been concluded by Klaman (1995) that the total gamut volume does not correlate well with visual appearance. We do, however, believe that the small gamut of the magnetographic device contributes to its results. Concerning the color quality evaluation through color difference measurements, we would like to remark that the numbers should be used with care. Although probably a good indicator for color quality, minimizing the average and maximum color differences does not guarantee optimal results in terms of perceived color image quality. For example, colors that are not in the evaluation target might be important. Another factor that limits this approach is gamut mapping (Morovic, 1998). Because of the differences in color gamut between different devices and technologies, a colorimetrically exact reproduction is rarely optimal. As we have mentioned earlier, the notion of color image quality is ultimately dened by what the customer wants. Since there does not yet to our knowledge exist measurement-based image quality models that adequately quanties quality in this sense, psychophysical experiments or panel tests is the method of choice. However, there are many critical factors that may limit the signicance of such an experiment, such as the number of observers, the relevance of the questions, and the choice of representative images. It should be mentioned that we have received indications that some of the presses were operated using suboptimal printing parameters, and that the printed samples thus do not give a good representation of the achievable quality. It would probably be much better to run the test prints, for each device model, independently at several locations, with different operators, but obviously this would add to the complexity of the study. It is also worth discussing the impact that this study has had, both within the participating companies and towards customers. It was not really our goal to inuence customers to buy one device instead of another. In our presentation at the Scandinavian conferences, we emphasized the uncertainties and unknowns in the process, and that there are much more than color image quality to con-

sider when buying a digital press, but that message was not really heard what came through was that one device is better than the other... In hindsight, we were probably nave not to realize the impact our study would have, and the participating companies probably did not realize this either.

5. Conclusion and Perspectives

Color image quality is of very high importance in a digital imaging device such as a digital press. For manufacturers and customers of such devices it is important to be able to quantify color image quality. However, to do so is not a trivial task, since ultimately, quality is dened as what the customer wants. Unfortunately, as of today there are no analytical techniques that can quantify color image quality in this context. It is therefore necessary to rely on experiments involving real observers. We claim that the notion of color image quality is ultimately tied to the preferences of customers and end users. Because of this, a very useful tool to quantify color image quality is psychophysical experiments involving a panel of human observers. However, it is clear that such experiments are relatively time consuming. Denitively, Yendrikhovskij (1999) hits the nail on the head when he states that most studies on image quality employ subjective assessment with only one goal to avoid it in the future. Therefore results from ongoing research toward analytical models for color image quality is eagerly anticipated. An example of such research is the development of metrics for color differences between complex images (CIE TC8-02, 2000, Imai et al., 2001). However, a device or algorithm that takes any image as input, and provides a number that perfectly quanties its color image quality as output, is still probably many, many years away.

Our thanks go to Peter Oll n and Aktuell Grask Infore mation AB for their support, and to our contacts at the four manufacturers for agreeing to spend time and consumables on our project.

Anderson, M., Motta, R., Chandrasekar, S., and Stokes, M. (1996). Proposal for a standard default color space for the internet sRGB. In Proceedings of IS&T and SIDs 4th Color Imaging Conference: Color Science, Systems and Applications, pages 238246, Scottsdale, Arizona. Updated version 1.10 can be found at http: // Bolanca, S., Mrvac, N., and Zjakic, I. (2001). Reproduction objectivity of four color and seven color conventional printing and digital offset printing technique. In

Advances in Color Reproduction Proceedings of the 28th International Iarigai Research Conference, chapter 4.4. GATF Press, Sewickley, PA. CIE 15.2 (1986). Colorimetry, volume 15.2 of CIE Publications. Central Bureau of the CIE, Vienna, Austria, 2 edition. CIE TC8-02 (2000). Methods to derive colour differences for images. Draft Version 0.4 of CIE Technical Report. Online: CIE TC8-05 (2001). Color encoding criteria. Draft Version 0.6 of CIE Technical Report. See http://www. den Engelsman, M. (2002). Is more colour also more beautiful? Pers, 14(21), 2021. (In Dutch: Is meer kleur ook mooier?) See Edinger, Jr., J. R. (2002). Scaling subjective impressions of quality. In Proceedings of NIP16: International Conference on Digital Printing Technologies, pages 377 382. Engeldrum, P. G. (2001). Psychometric scaling: Avoiding the pitfalls and hazards. In Proc. IS&Ts 2001 PICS Conference, pages 101107, Montr al, Qu b c, e e e Canada. Farup, I. and Hardeberg, J. Y. (2002). Visualization and interactive manipulation of color gamuts. In IS&T and SIDs 10th Color Imaging Conference: Color Science, Systems and Applications, Scottsdale, Arizona. To appear. Grice, J. and Allebach, J. P. (1999). The print quality toolkit: An integrated print quality assessment tool. Journal of Imaging Science and Technology, 43(2), 187199. Hardeberg, J. Y. (2002). Color image quality for multifunction peripherals. In Proc. IS&Ts 2002 PICS Conference, pages 7681, Portland, Oregon. Hardeberg, J. Y. and Farup, I. (2002). Interactive color gamut mapping. In International Printing and Graphic Arts Conference, Bordeaux, France. To appear. ICC.1:2001-12 (2001). File format for color proles. The International Color Consortium. Version ICC.1:200112, See IEC 61966-2.1 (1999). Colour measurement and management in multimedia systems and equipment, Part 21: Colour management - Default RGB colour space sRGB. IEC Publication, 61966-2.1. Imai, F. H., Tsumura, N., and Miyake, Y. (2001). Perceptual color difference metric for complex images based on Mahalanobis distance. Journal of Electronic Imaging, 10(2), 385393. Jung, J., Geelen, R., and Vandenbroucke, D. (2001). The virtual image chain: A powerful tool for the evaluation of the perceived image quality as a function of imaging system parameters. In Proc. IS&Ts 2001 PICS Conference, Montr al, Qu b c, Canada. e e e Kane, P. J., Bouk, T. F., Burns, P. D., and Thompson, A. D. (2000). Quantication of banding, streaking and grain

in at eld images. In Proc. IS&Ts 2000 PICS Conference, Portland, Oregon. Kipphan, H., editor (2001). Handbook of Printed Media. Springer-Verlag, Berlin. Klaman, M. (1995). A test model for proong systems. In Proceedings of 1st joint TAGA/IARIGAI Technical conference, Paris, France. Klaman, M. (2001). Color rendering aspects in digital printing. In TAGA Proceedings. Lindberg, S., Donderi, D. C., and Pauler, N. (2001). Image quality of four digital printing methods compared to offset and exographic printing. In Advances in Color Reproduction Proceedings of the 28th International Iarigai Research Conference, chapter 4.1. GATF Press, Sewickley, PA. McDonald, R. and Smith, K. J. (1995). CIE94 a new colour-difference formula. J. Soc. Dyers Col., 111, 376379. Microsoft Corporation (2001). Windows Color Quality Specications for Printer OEMs. Part of Microsoft Hardware Quality Labs (WHQL)s Windows Color Quality Test Kit. hwdev/tech/color/ColorTest.asp. Morovic, J. (1998). To Develop a Universal Gamut Mapping Algorithm. Ph.D. thesis, Colour & Imaging Institute, University of Derby. Norberg, O., Westin, P., Lindberg, S., Klaman, M., and Eidenvall, L. (2002). Comparison of print quality between digital and traditional technologies. In Proceedings of DPP2001: International Conference on Digital Production Printing and Industrial Applications, pages 380385. Rasmussen, D. R., Crean, P. A., Nakaya, F., Sato, M., and Dalal, E. N. (1998). Image quality metrics: Applications and requirements. In Proc. IS&Ts 1998 PICS Conference, pages 216220, Portland, Oregon. Skarsb, S. E. and Hardeberg, J. Y. (2002). Digital presses how well do they print? Aktuell Grask Information, (248). (In Swedish: Digitalpressarna hur bra trycker de egentligen?). Stokes, M. (1998). The impact of color management terminology on image quality. In Proc. IS&Ts 1998 PICS Conference, pages 174178, Portland, Oregon. T pfer, K. and Cookingham, R. (2000). The quantitative o aspects of color quality rendering for memory colors. In Proc. IS&Ts 2000 PICS Conference, Portland, Oregon. Wyszecki, G. and Stiles, W. S. (1982). Color Science: Concepts and Methods, Quantitative Data and Formulae. John Wiley & Sons, New York, 2 edition. Yendrikhovskij, S. N. (1999). Image quality: Between science and ction. In Proc. IS&Ts 1999 PICS Conference, pages 173178, Savannah, Georgia.