Received June 13, 2016, accepted July 4, 2016, date of publication July 18, 2016, date of current version

August 15, 2016.

Digital Object Identifier 10.1109/ACCESS.2016.2588283

Display Gamut Metrology Using

Chromaticity Diagram
KENICHIRO MASAOKA
Japan Broadcasting Corporation (NHK), Tokyo 157-8510, Japan (masaoka.k-gm@nhk.or.jp)

ABSTRACT With the introduction of the International Telecommunication Union—Radiocommunication

Sector (ITU-R) Recommendation BT.2020 (Rec. 2020) color space for ultrahigh definition television, the
development of wide-gamut displays has become a high priority. To measure the relative gamut sizes
of wide-gamut displays, it is herein proposed that the International Commission on Illumination (CIE)
1931 xy chromaticity diagram be used as a practical single metric rather than the nominally perceptually
uniform CIE 1976 u0 v0 chromaticity diagram because of the high correlations between the Rec. 2020
gamut area-coverage ratios in the xy diagram and the volume-coverage ratios in the CIE 1976 L ∗ a∗ b∗ and
CIE 1976 L ∗ u∗ v∗ color spaces. This paper explains why using the perceptually nonuniform xy diagram is
counterintuitively appropriate to measure display gamut sizes by presenting a quantitative analysis of the
gamut coverage in terms of hue. The argument is made that performing planimetry in the u0 v0 diagram,
whose perceptual uniformity is valid only when the luminance is constant, does not yield accurate display
gamut sizes, because the magenta region area is overestimated and the yellow region area is underestimated.
Furthermore, some limitations on the use of planimetry for measuring display gamut sizes of multiprimary
displays are discussed, and the color light output standard is proposed as a complement to practical
planimetry using the xy diagram.

INDEX TERMS Chromaticity diagram, display gamut size, planimetry, Recommendation ITU-R BT.2020,
ultrahigh definition television, wide color gamut.

I. INTRODUCTION and Television Engineers (SMPTE) Recommended Practice

The wide-gamut system colorimetry for ultrahigh defini-
tion television (UHDTV) was designed based on color
encoding efficiency, program quality management, target
gamut achievement, and gamut mapping efficiency require-
ments [1]. To enable high color encoding efficiency
and program quality management, physically realizable,
highly saturated red, green, and blue (RGB) primaries
were selected in International Telecommunication Union—
Radiocommunication Sector (ITU-R) Recommendation
BT.2020 (Rec. 2020) [2] instead of additional or unreal
primaries. Program quality management requires each color
to be reproducible on a reference display so that broadcasters
can monitor and control the color reproduction. Thus, to
maximize the color gamut and fulfill the aforementioned
requirements, monochromatic RGB primaries were chosen
in Rec. 2020, which encompasses the color spaces of the
major standard system colorimetries: ITU-R Recommen-
dation BT.709 (Rec. 709) (for high definition television)
[3], Adobe RGB [4] (as a de facto standard in commercial
photography and graphic arts), and Society of Motion Picture
431-2:2011 [5] (i.e., the DCI-P3 standard for the reference
digital cinema projector). Fig. 1 shows the chromaticities of
the RGB primary sets and Pointer’s colors [6] (representing
the maximum gamut of real object colors under illuminant C)
transformed to those under illuminant D65 using the CAT02
chromatic adaptation transform [7] in the International Com-
mission on Illumination (CIE) 1931 xy and CIE 1976 u0 v0
chromaticity diagrams. The Adobe RGB red and green pri-
maries are identical to the Rec. 709 red primary and (obsolete)
National Television System Committee (NTSC) 1953 green
primary, respectively, whereas the Adobe RGB and DCI-P3
blue primaries are identical to the Rec. 709 blue primary.
With the introduction of the Rec. 2020 color space,
the development of direct-view wide-gamut displays has
become a high priority, even though it is a major challenge.
While liquid crystal displays (LCDs) are considered promis-
ing as UHDTV displays and laser-backlit LCDs attain the
Rec. 2020 color space [8], nonmonochromatic light sources
may also be used due to their cost and performance benefits.
Many displays that support the Rec. 2020 color space are

2169-3536
2016 IEEE. Translations and content mining are permitted for academic research only.
3878 Personal use is also permitted, but republication/redistribution requires IEEE permission. VOLUME 4, 2016
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
K. Masaoka: Display Gamut Metrology Using Chromaticity Diagram
FIGURE 1. Chromaticities of RGB primary sets of Rec. 2020, Adobe RGB,
DCI-P3, and Rec. 709 color spaces and chromaticities of Pointer’s colors
under illuminant D65 in (a) CIE 1931 xy and (b) CIE 1976 u0 v 0 chromaticity
diagrams.
commercially available, although their gamut coverages vary.
It has been reported that quantum-dot-based LCDs can
achieve more than 90% area coverage of the Rec. 2020 color
space in both the xy and u0 v0 diagrams [9]. However, most
such displays can approximately achieve the Adobe RGB
or DCI-P3 color space, which cover only about 70% of the
Rec. 2020 color space.
In the development of wide-gamut displays of varying
gamut sizes, gamut size measurement has become increas-
ingly important, and an appropriate metric is urgently needed.
A popular metric involves comparing the area of the triangle
formed by connecting the RGB primaries (RGB triangle) in
the xy or u0 v0 chromaticity diagram to the area of the RGB
triangle of a standard RGB color space, such as NTSC, Adobe
RGB, DCI-P3, or (recently) Rec. 2020. The u0 v0 diagram,
VOLUME 4, 2016
FIGURE 2. MacAdam ellipses and Rec. 2020 RGB triangle in (a) xy and
(b) u0 v 0 diagrams. Axes of plotted ellipses are 10 times their actual
lengths.
which is a projective transformation of the xy diagram, yields
perceptually more uniform color spacing and has gained
popularity for use in planimetry. Fig. 2 shows MacAdam
ellipses, whose contours represent the just-noticeable differ-
ences (JNDs) of chromaticity in the xy and u0 v0 diagrams [10].
The ellipses in the u0 v0 diagram look more rounded than those
in the xy diagram. However, from the point of view of color
science, any reference to area coverage in a chromaticity
diagram is inappropriate because a color gamut is consid-
ered to form a solid in a three-dimensional perceptual color
space inherently due to the trichromatic nature of human
vision, in which a color is represented in cylindrical coordi-
nates (chroma, hue, and lightness) or Cartesian coordinates
(two opponent color dimensions and lightness). Fig. 3 shows
Pointer’s colors (under illuminant D65) and the Rec. 709
and Rec. 2020 gamuts in the CIE xyY (upper-case Y denotes
relative luminance) and CIE L ∗ a∗ b∗ (CIELAB) color spaces.
The mean JND data form ellipsoids of various sizes and
orientations in the xyY color space, leading to the creation
of the CIELAB color space, which is nominally perceptually
uniform with spherical JND solids assumed in the space.
In the xyY color space as well as the u0 v0 Y color space
(not shown in Fig. 3), each gamut has a flat triangular base
(i.e., an RGB triangle) with three flat vertical walls that
come together at a roof consisting of three warped plates,
creating an enclosed volume. The area of the RGB triangle
base does not accurately represent the gamut volume in the
xyY and u0 v0 Y color spaces because the solid is neither a
triangular prism nor a triangular pyramid. As shown in Fig. 3,
the base RGB triangle begins to lack the blue corner at a low
relative luminance. Therefore, the base RGB triangle area is
proportional to the gamut volume in the xyY or u0 v0 Y color
space only up to this low relative luminance. In addition,
near-black colors sweep the xy plane in the xyY color space,
making the Pointer’s gamut teardrop-shaped, whereas black
converges to a point in the CIELAB color space. Regard-
less of these fundamental differences, display manufacturers
define relative display gamut sizes in two dimensions because
planimetry is simple in terms of calculation and historically
has yielded continuous metrics.
3879

FIGURE 3. Pointer’s colors under illuminant D65 (colored dots) and
gamuts (translucent) corresponding to (a) Rec. 709 and (b) Rec. 2020 in
CIE xyY color space and (c) Rec. 709 and (d) Rec. 2020 in CIELAB color
space.
There is a major problem with this pragmatic approach
related to the discrepancy between the areas in the
xy and u0 v0 chromaticity diagrams. Display manufactur-
ers compare the area ratios of their displays’ RGB tri-
angles with the area of a standard RGB triangle in the
xy or u0 v0 chromaticity diagram to determine the relative
gamut sizes of the displays. In reality, manufacturers typ-
ically neglect to mention which of the two diagrams the
gamut sizes that they report are based on, instead tend-
ing to use the larger of the two resulting sizes. Another
problem is the ambiguous practice of using two different
ratios: the ‘‘area-size ratio’’ (Adisplay / Astandard × 100%),
which is the ratio of a display’s RGB triangle area (Adisplay )
to the standard RGB triangle area (Astandard ), and the ‘‘area-
coverage ratio’’ (Adisplay ∩ Astandard / Astandard × 100%),
which is the ratio of the area of the polygon-shaped over-
lap region between the display and standard RGB triangles
(Adisplay ∩ Astandard ) to Astandard . If a primary color is outside
the target color space, the area-size ratio becomes larger than
the area-coverage ratio and can be more than 100%.
Irrespective of volumetry, planimetry, the gamut-size ratio,
and the gamut-coverage ratio, it is reasonable to use a single
metric to measure display gamut sizes. Before addressing
the validity of volumetry or planimetry, it is necessary to
define the target color space. Rec. 2020 is the most promising
wide-gamut color space, in that its RGB color space covers
the major standard color spaces and most real object colors,
and it is also the standard gamut for UHDTV, making it the
most suitable target gamut for wide-gamut displays. From
the point of view of source color reproduction accuracy,
3880
K. Masaoka: Display Gamut Metrology Using Chromaticity Diagram
FIGURE 4. Approaches used to obtain unifying relative display gamut size
metrics.
the gamut-coverage ratio rather than the gamut-size ratio
is valid. Given this background information, the two
approaches shown in Fig. 4 should be considered. One is a
pragmatic approach to planimetry in a single chromaticity
diagram. The other is a scientific approach in which the
gamut volume is calculated in a color appearance space.
Masaoka and Nishida [11] used a heuristic approach to com-
pare the Rec. 2020 area-coverage ratios in the xy and u0 v0
diagrams to the Rec. 2020 volume-coverage ratios in the
CIELAB, CIE 1976 L ∗ u∗ v∗ (CIELUV), and CIECAM02
(the latest standardized color appearance model) [7] color
spaces. They proposed an approach based on the xy dia-
gram rather than the perceptually uniform u0 v0 diagram to
determine the gamut metric, since their simulation results
revealed higher correlations between the Rec. 2020 area-
coverage ratios in the xy diagram and the Rec. 2020 volume-
coverage ratios in each of the color appearance spaces. Some
new color appearance models and color difference equations
(e.g., CAM02-UCS [12], CIE94 [13], and CIEDE2000 [14])
are thought to yield gamut sizes more accurate than those
obtainable using CIELAB or CIELUV. However, their gamut
predictions differ significantly from one another, in spite of
their computational complexity, and it is inordinately difficult
to determine the true values [15], [16]. Therefore, it is advis-
able to use an approach more reasonable than attempting to
determine the true volume.
However, it seems counterintuitive to use the xy diagram
rather than the perceptually uniform u0 v0 diagram to perform
planimetry. Herein, the use of the xy diagram to measure
relative display gamut sizes is validated by presenting a quan-
titative analysis of gamut sizes divided into three hue regions:
cyan, magenta, and yellow (CMY).
II. ANALYSIS
A. SAMPLING WIDE-GAMUT RGB PRIMARIES
The same wide-gamut RGB primary sets as in [11] were
used. Each display primary is within the area enclosed by
the extended sides of the Rec. 709 RGB triangle, horseshoe-
shaped spectrum locus, and purple boundary in a CIE
chromaticity diagram, as depicted in Fig. 5. Any RGB
primary set can enclose the Rec. 709 RGB triangle. Fig. 6
shows the sampled RGB primaries in enlarged sections
of Fig. 5. The RGB primaries were sampled as evenly as
VOLUME 4, 2016
K. Masaoka: Display Gamut Metrology Using Chromaticity Diagram

FIGURE 7. Rec. 2020 CMY regions in (a) xy and (b) u0 v 0 diagrams.

Markers denote standard RGB primary sets, as in Fig. 1 legend.

The Rec. 2020 area-coverage ratios were estimated in the

FIGURE 5. RGB primaries sampled to cover Rec. 709 color space. CMY hue regions separately. Fig. 7 shows the Rec. 2020
CMY regions in the xy and u0 v0 diagrams. The Rec. 2020
RGB triangles in the two diagrams are each divided into
three regions colored in CMY, where the straight borderlines
connect the RGB chromaticity points with the reference white
point (D65). To determine the Rec. 2020 cyan, magenta,
or yellow area-coverage ratio (ARGB ∩ A2020(C/M/Y) /
A2020 × 100%), where the subscript RGB denotes a sampled
RGB triangle and the subscript 2020(C/M/Y) denotes the
cyan, magenta, or yellow region of the Rec. 2020 RGB trian-
gle, the coordinates of the overlap polygon and the area were
computed using the Sutherland–Hodgman algorithm [17] and
FIGURE 6. Enlarged sections of Fig. 5 containing sampled RGB primaries. the shoelace formula [18], respectively.
Gray circle, green square, brown diamond, and black circle markers
denote Rec. 2020, Adobe RGB, DCI-P3, and Rec. 709 RGB primary sets,
respectively, as in Fig. 1 legend. C. COMPUTATION OF VOLUME-COVERAGE RATIOS
IN CIELAB AND CIELUV COLOR SPACES
possible from each area in the uniform u0 v0 diagram. A total The Rec. 2020 cyan, magenta, or yellow volume-coverage
of 24 red, 33 green, and 15 blue primaries were sampled ratio (VRGB ∩ V2020(C/M/Y) / V2020 × 100%) is the ratio of
(including the RGB primaries for Rec. 709, Adobe RGB, the volume of the solid overlap region between the gamuts
DCI-P3, and Rec. 2020), yielding 11,880 different RGB pri- of an RGB primary set and a Rec. 2020 cyan, magenta, or
mary sets all together. yellow region (VRGB ∩ V2020(C/M/Y) ) to the volume of the
Rec. 2020 gamut (V2020 ) in a color appearance space. The
B. COMPUTATION OF AREA-COVERAGE RATIOS volume-coverage ratios in the CIELAB and CIELUV color
IN xy AND u0 v0 DIAGRAMS spaces were calculated in this study and are discussed herein.
Chromaticity values x and y are computed by normalizing Both CIELAB and CIELUV are primitive color appear-
CIE 1931 XYZ tristimulus values: ance models standardized to achieve colors that are percep-
tually uniform. CIELAB is used in the colorant industry,
x = X /(X + Y + Z ) and it has become universally applied for color specification.
y = Y /(X + Y + Z ). (1) CIELUV which is based on the u0 v0 diagram, was once pop-
ular in the television industry, although it has since become
In this projective transformation, any additive mixture of two obsolete [19]. The CIELAB color space is expressed by the
sets of tristimulus values follows a straight line in the xy plane. following coordinates along three orthogonal axes:
Chromaticity values u0 and v0 are determined using
L ∗ = 116 f (Y /Yn ) − 16
u0 = 4X /(X + 15Y + 3Z )
a∗ = 500 (f (X /Xn ) − f (Y /Yn ))
v0 = 9Y /(X + 15Y + 3Z ). (2)
b∗ = 200 (f (Y /Yn ) − f (Z /Zn )), (4)
Equivalently, u0 and v0 can be computed based on x and y:
where L ∗ indicates lightness; a∗ and b∗ are the coordinates
u0 = 4x/(3 − 2x + 12y) in the color-opponent dimensions; X , Y , and Z denote the
v0 = 9y/(3 − 2x + 12y). (3) CIE tristimulus values of a color under consideration; Xn , Yn ,

VOLUME 4, 2016 3881

 K. Masaoka: Display Gamut Metrology Using Chromaticity Diagram

and Zn represent the tristimulus values of the reference white;

and
f (Y /Yn ) = (Y /Yn )1/3 for Y /Yn > 0.008856
f (Y /Yn ) = 7.787Y /Yn + 16/116 otherwise. (5)
The functions f (X /Xn ) and f (Z /Zn ) can be obtained in the
same manner. The CIELUV color space is also expressed
using the coordinates along the three orthogonal axes:
L ∗ = 116 f (Y /Yn )1/3 − 16
u∗ = 13L ∗ (u0 − u0n )
v∗ = 13L ∗ (v0 − v0n ), (6)
where L ∗ has the same definition as in the CIELAB color
space, u∗ and v∗ are the coordinates in the color-opponent
dimensions, and u0n and v0n are the u0 v0 chromaticity coordi-
nates of the reference white point. The CIELAB chroma Cab ∗
∗
and CIELUV chroma Cuv are expressed as follows:
∗
Cab = (a∗2 + b∗2 )1/2
∗
Cuv = (u∗2 + v∗2 )1/2 . (7)
Furthermore, the CIELAB hue angle hab and CIELUV hue
angle huv are expressed as follows:
hab = tan−1 (b∗ /a∗ )
huv = tan−1 (v∗ /u∗ ). (8)
To estimate the gamut volume for each sampled RGB
primary set, 100 gamut boundary polygons in the xyY color
space were first obtained at the relative luminances Y corre-
sponding to lightness values L ∗ of 0.5, 1.5, . . . , and 99.5. The
normalized RGB levels of each vertex of each polygon may
have any of the following sets of values: [n 0 0], [1 n 0], [1 1
n], [n 1 0], [0 n 0], [0 1 n], [n 1 1], [0 n 1], [0 0 n], [n 0 1],
[1 n 1], or [1 0 n], where 0 ≤ n ≤ 1. The xy coordinates of the
vertices of each polygon having three, four, or five vertices
were converted into u0 v0 and u∗ v∗ coordinates, forming new
polygons with the same numbers of vertices. To obtain the
constant-L ∗ loci in the CIELAB color space, in which the
polygon edges are curved, 100 points between each pair of
adjacent xy vertices were interpolated in advance for each
polygon and then were converted into a∗ b∗ coordinates. Fig. 8
shows the constant-L ∗ contour plots of the Rec. 2020 gamut FIGURE 8. Constant-L∗ contour plots of Rec. 2020 CMY gamuts with
in the xy, a∗ b∗ , and u∗ v∗ diagrams. To calculate the Rec. 2020 L∗ = 0.5, 3.5, . . . , 99.5 (instead of L∗ increments of 1 to improve
legibility) in (a) xy , (b) a∗ b∗ , and (c) u∗ v ∗ diagrams.
CMY volume-coverage ratios in each color space, the coor-
dinates of the vertices of the polygonal overlap area (which
contains hundreds of vertices in the a∗ b∗ diagram) and its size the Rec. 2020 CMY volume-coverage ratios in the CIELAB
at each lightness value were computed using the Sutherland– and CIELUV color spaces. The RGB levels for each dot color
Hodgman algorithm and the shoelace formula, respectively, are proportional to the differences between the u0 v0 saturation
and the volume was approximated by the sum of the areas of values for the Rec. 709 RGB primaries and the sampled RGB
the 100 constant-L ∗ loci. primaries, showing the saturation balance for the sampled
RGB primary set. The black circle marker, having the lowest
D. COMPARISON OF AREA-COVERAGE AND coverage ratio, denotes the Rec. 709 RGB primary set. The
VOLUME-COVERAGE RATIOS light gray circle marker, with the highest coverage ratio,
Figs. 9 and 10 present the comparisons between the Rec. 2020 denotes the Rec. 2020 RGB primary set. The green dots indi-
CMY area-coverage ratios in the xy and u0 v0 diagrams and cate that the green primaries for the sampled RGB primary

3882 VOLUME 4, 2016

K. Masaoka: Display Gamut Metrology Using Chromaticity Diagram

FIGURE 9. Comparison between Rec. 2020 volume-coverage ratios in CIELAB color space and Rec. 2020 area-coverage
ratios for (a) cyan, (b) magenta, and (c) yellow regions in xy diagram and (d) cyan, (e) magenta, and (f) yellow regions
in u0 v 0 diagram. Each dot color corresponds to balance of saturation for sampled RGB primary set (see text). Gray
circle, green square, brown diamond, and black circle markers denote Rec. 2020, Adobe RGB, DCI-P3, and Rec. 709 RGB
primary sets, respectively, as in Fig. 1 legend, and green and magenta circle markers denote R709 G2020 B709 and
R2020 G709 B2020 primary sets, respectively.

sets are relatively saturated. The magenta dots demonstrate
that the red and blue primaries are relatively saturated. The
green and magenta circle markers denote the R709 G2020 B709
and R2020 G709 B2020 primary sets, respectively (the subscript
numbers indicate the ITU-R recommendations used to define
the RGB primaries). For each hue region, it is evident that
the correlations of the area-coverage ratios in the xy and u0 v0
diagrams with the volume-coverage ratios in each color space
are comparable. For each chromaticity diagram and each
color space, the correlation between the Rec. 2020 cyan area-
coverage and volume-coverage ratios is quite high. For the
Rec. 2020 magenta region, on the other hand, the volume-
coverage ratios spread out. Moreover, for the u0 v0 diagram, the
magenta area-coverage ratios have large values with a wide
range, whereas the yellow area-coverage ratios have small
values with a narrow range, which decreases the correlation
between the Rec. 2020 area-coverage and volume-coverage
ratios. Figs. 11 and 12 present the comparisons between the
Rec. 2020 area-coverage ratios in the xy and u0 v0 diagrams
and the Rec. 2020 volume-coverage ratios in the CIELAB
and CIELUV color spaces for the magenta and yellow regions
combined as well as for the CMY regions combined. The
area-coverage ratios in the xy diagram for the magenta and
yellow regions combined are much better correlated with
the volume-coverage ratios for the combined region in each
color space than are those in the u0 v0 diagram. As a result,
the volume-coverage ratios in the CMY regions combined
are more closely correlated with the area-coverage ratios in
the xy diagram than with those in the u0 v0 diagram, even in
the CIELUV color space, in spite of the fact that the u0 v0
coordinates are precursors of the u∗ v∗ coordinates, as in (6).
Note that the results in the CMY regions combined that were
computed by summing the CMY coverage ratios completely
match Masaoka and Nishida’s results [11].
III. DISCUSSION
The analysis clearly shows that the xy diagram is more
suitable than the perceptually uniform u0 v0 diagram for plani-
metric measurements of relative display gamut sizes. The
area-coverage ratios in the u0 v0 diagram are highly incon-
sistent compared to the area-coverage ratios in the xy dia-
gram and the volume-coverage ratios in the color appearance
spaces due to the overestimation of the magenta region and
the underestimation of the yellow region determined using
the u0 v0 diagram.
The observation that the Rec. 2020 area-coverage ratios
in the xy diagram are better correlated than those in the u0 v0
diagram with the Rec. 2020 volume-coverage ratios in each
of the color appearance spaces introduces the question of
why the xy diagram is more appropriate than the u0 v0 diagram

VOLUME 4, 2016 3883


FIGURE 10. Comparison between Rec. 2020 volume-coverage ratios in CIELUV color space and Rec. 2020
area-coverage ratios for (a) cyan, (b) magenta, and (c) yellow regions in xy diagram and (d) cyan, (e) magenta,
and (f) yellow regions in u0 v 0 diagram. Markers denote standard RGB primary sets, as shown in Fig. 9.
in spite of the perceptual uniformity of the u0 v0 diagram.
The key to the answer is the factor of L ∗ in the u∗ and v∗
formulas shown in (6). The chroma Cuv ∗ is proportional to
∗
L , and hence, colors with high luminances, such as green
and yellow colors on a display gamut surface, have high
chromas. Fig. 13 shows the chromaticity distributions of
Pointer’s colors (under illuminant D65) in the a∗ b∗ and u∗ v∗
diagrams. The distribution in the a∗ b∗ diagram appears sim-
ilar to that in the u∗ v∗ diagram if the a∗ b∗ diagram is rotated
clockwise by approximately 45◦ . The distribution in the u∗ v∗
diagram appears more similar to that in the xy diagram than
the u0 v0 diagram, as shown in Fig. 1. In the u0 v0 diagram,
the distributions of green and yellow are shrunken and that
of magenta is expanded. Thus, the perceptual uniformity
of the u0 v0 diagram does not necessarily make it appro-
priate for use in the planimetric determination of gamut
volumes. In fact, the volume-coverage ratios of Pointer’s
gamut in the CIELAB, CIELUV, and CIECAM02 color
spaces are better correlated with the Rec. 2020 area-coverage
ratios for the sampled RGB primary sets in the xy diagram
(r = 0.925, 0.934, and 0.913 for CIELAB, CIELUV, and
CIECAM02, respectively) than with those in the u0 v0 dia-
gram (r = 0.669, 0.735, and 0.675 for CIELAB, CIELUV,
and CIECAM02, respectively) [11]. The results validate the
practicality of using planimetry in the xy diagram to deter-
mine display gamut sizes for TV applications in terms of
visual importance.
3884
K. Masaoka: Display Gamut Metrology Using Chromaticity Diagram
Furthermore, the Adobe RGB and DCI-P3 volume-
coverage ratios in the CIELAB color space are better corre-
lated with the Adobe RGB and DCI-P3 area-coverage ratios
in the xy diagram (r = 0.998 and 0.990 for Adobe RGB
and DCI-P3, respectively) than with those in the u0 v0 dia-
gram (r = 0.988 and 0.749 for Adobe RGB and DCI-P3,
respectively) [20], suggesting the appropriateness of using
the xy diagram regardless of the target gamut. Although
Rec. 2020 is the most suitable target color gamut, as men-
tioned in the Introduction, the Adobe RGB and DCI-P3 color
spaces will remain important target gamuts for computer
monitors, mobile displays, and organic light-emitting diode
devices.
The nonuniform xy diagram is used only as an intermediate
space to approximate display gamut volumes. The enlarge-
ment of the yellow and green regions of the xy diagram coin-
cidently enhances its suitability as a metric. While comparing
different colors with the same luminance levels, the percep-
tually uniform u0 v0 diagram is suitable. For example, the
chromaticity tolerance of a display primary and screen color
uniformity would reasonably be expressed in the
u0 v0 diagram. However, it is incorrect to conclude that the
u0 v0 diagram is usable in planimetry for measuring display
gamuts based on its perceptual uniformity, although many
scientists and engineers draw this conclusion. In the Infor-
mation Display Measurements Standard (IDMS) [21] devel-
oped by the International Committee for Display Metrology,
VOLUME 4, 2016
K. Masaoka: Display Gamut Metrology Using Chromaticity Diagram
FIGURE 11. Comparison between Rec. 2020 volume-coverage ratios in
CIELAB color space and Rec. 2020 area-coverage ratios in xy diagram for
(a) magenta and yellow regions combined and (b) CMY regions combined
and in u0 v 0 diagram for (c) magenta and yellow regions combined and (d)
CMY regions combined. Markers denote standard RGB primary sets, as
shown in Fig. 9.
which is part of the Society for Information Display (SID),
in cooperation with the Video Electronics Standards Associ-
ation (VESA), the following statement is written:
. . . we strongly encourage people to abandon the
use of the 1931 CIE color diagram for determining
the color gamut because it does not have a Euclid-
ian metric defined in it. The 1976 CIE (u0 ,v0 ) color
diagram should be used instead. Unfortunately,
many continue to use the (x,y) chromaticity values
and the 1931 diagram for gamut areas.
This statement is incorrect. Poynton’s book on video tech-
nology directly points out the misleading nature of this
argument [22]:
Perceptual differences in a [u0 , v0 ] diagram are
dependent upon luminance, but that fact is not
evident from the diagram: Be careful not to draw
strong conclusions from the diagram.
Even if any new color appearance models are developed in
the future, it is not expected that the appropriateness of using
the xy diagram rather than the u0 v0 diagram to determine the
metric would be reversed, as long as the dependence of the
chroma on the lightness is properly taken into account.
To illustrate the cause of the low correlations between
the Rec. 2020 magenta area-coverage ratios and volume-
coverage ratios, Fig. 14 shows the gamut boundaries of
Rec. 709, R2020 G709 B2020 , R709 G2020 B709 , and Rec. 2020
in the L ∗ a∗ plane (hab = 0◦ and 180◦ ). It is evident that a
VOLUME 4, 2016
FIGURE 12. Comparison between Rec. 2020 volume-coverage ratios in
CIELUV color space and Rec. 2020 area-coverage ratios in xy diagram for
(a) magenta and yellow regions combined and (b) CMY regions combined
and in u0 v 0 diagram for (c) magenta and yellow regions combined and (d)
CMY regions combined. Markers denote standard RGB primary sets, as
shown in Fig. 9.
FIGURE 13. Chromaticities of Pointer’s colors under illuminant D65
in (a) a∗ b∗ and (b) u∗ v ∗ diagrams.
saturated green primary boosts the magenta volume because
brighter red and blue primaries are needed to set the white
point to D65. In spite of the limitation that the effect of a
saturated green primary boosting the gamut in magenta is not
considered in the investigated planimetric method, evaluating
the Rec. 2020 area coverage of the CMY regions separately
will be practical and helpful in designing wide-gamut dis-
plays for the Rec. 2020 coverage in terms of hue.
The validity of this approach has been demonstrated
only for three-primary color displays. For multiprimary
color displays with more than three primary colors, this
approach may not be suitable. In particular, white or any
3885

FIGURE 14. Gamut boundaries of Rec. 709, R2020 G709 B2020 ,
R709 G2020 B709 , and Rec. 2020 in L∗ a∗ plane (hab = 0◦ and 180◦ ).
other primary inside the RGB triangle is fundamentally
different from cyan, magenta, yellow, or any other pri-
mary outside the RGB triangle, as is easy to see conceptu-
ally. A white primary used with RGB primaries increases
the brightness without increasing the gamut [23]. In this
case, planimetric gamut measurements using a chromatic-
ity diagram are insufficient. A realistic complement is the
color light output (CLO) metric specified in the IDMS.
The CLO metric is a specification that provides informa-
tion about color brightness based on the total luminous flux
(for front-projection displays) or luminance (for rear-
projection and emissive displays) from the full-on pure RGB
colors of a display, and it is typically provided in addition
to the white light output (WLO) metric, which only mea-
sures white light. With the CLO and WLO metrics, using
planimetry in the xy diagram to measure display gamut sizes
is practical.
IV. CONCLUSION
Based on the high correlation between the Rec. 2020 volume-
coverage ratios in different color appearance spaces and the
Rec. 2020 area-coverage ratios in the xy chromaticity dia-
gram, the use of the xy diagram is recommended as a reason-
able metric for calculating Rec. 2020 area-coverage ratios.
The perceptual uniformity of the u0 v0 chromaticity diagram
is innately valid only when the luminance is constant. When
performing planimetry to measure relative display gamuts
in the u0 v0 diagram, the magenta region is overestimated,
whereas the yellow region is underestimated. Thus, it can-
not be concluded that the u0 v0 diagram is valid for use in
planimetry based on its perceptual uniformity. The CLO
standard is proposed as a complement to practical planimetry
using the xy diagram. It is expected that using the xy diagram
rather than the u0 v0 diagram as a single metric will facili-
tate the unbiased development of wide-gamut displays and
the standardization of proper metrics for measuring display
gamut sizes.
REFERENCES
[1] K. Masaoka, Y. Nishida, M. Sugawara, and E. Nakasu, ‘‘Design of pri-
maries for a wide-gamut television colorimetry,’’ IEEE Trans. Broadcast.,
vol. 56, no. 4, pp. 452–457, Apr. 2010.
3886
K. Masaoka: Display Gamut Metrology Using Chromaticity Diagram
[2] Parameter Values for Ultra-High Definition Television Systems for Pro-
duction and International Programme Exchange, document Rec. ITU-R
BT.2020-2, ITU, Geneva, Switzerland, 2015.
[3] Parameter Values for the HDTV Standards for Production and Inter-
national Programme Exchange, document Rec. ITU-R BT.709-6, ITU,
Geneva, Switzerland, 2015.
[4] Adobe RGB (1998) Color Image Encoding, Adobe Syst., Inc., San Jose,
CA, USA, 2005.
[5] D-Cinema Quality—Reference Projector and Environment,
document SMPTE RP 431-2, Society of Motion Picture and Television
Engineers, New York, NY, USA, 2011.
[6] M. R. Pointer, ‘‘The gamut of real surface colours,’’ Color Res. Appl.,
vol. 5, no. 3, pp. 145–155, 1980.
[7] A Colour Appearance Model for Colour Management Systems:
CIECAM02, document CIE 159, International Commission on
Illumination, CIE, Vienna, Austria, 2004.
[8] E. Niikura et al., ‘‘Development of RGB laser backlit liquid crystal
display,’’ in Proc. 22nd Int. Display Workshops, Otsu, Japan, 2015,
pp. 1096–1099.
[9] J. M. Hillis, J. Thielen, J. Tibbits, A. Lathrop, D. Lamb, and
J. Van Derlofske, ‘‘Closing in on Rec. 2020: How close is close enough?’’
in SID Symp. Dig. Tech., 2015, vol. 46. no. 1, pp. 223–226.
[10] G. Wyszecki and W. S. Stiles, ‘‘Visual equivalence and visual matching,’’
in Color Science: Concepts and Methods, Quantitative Data and Formula,
vol. 2, 2nd ed. Chichester, U.K.: Wiley, 2000, ch. 5, sec. 4.1, p. 309.
[11] K. Masaoka and Y. Nishida, ‘‘Metric of color-space coverage for wide-
gamut displays,’’ Opt. Exp., vol. 23, no. 6, pp. 7802–7808, 2015.
[12] M. R. Luo, G. Cui, and C. Li, ‘‘Uniform colour spaces based on
CIECAM02 colour appearance model,’’ Color Res. Appl., vol. 31, no. 4,
pp. 320–330, 2006.
[13] Industrial Colour-Difference Evaluation, document CIE 116, International
Commission on Illumination, CIE, Vienna, Austria, 1995.
[14] M. R. Luo, G. Cui, and B. Rigg, ‘‘The development of the CIE 2000
colour-difference formula: CIEDE2000,’’ Color Res. Appl., vol. 26, no. 5,
pp. 340–350, 2001.
[15] K. Masaoka, R. S. Berns, M. D. Fairchild, and F. M. Abed, ‘‘Number of
discernible object colors is a conundrum,’’ J. Opt. Soc. Amer. A, vol. 30,
no. 2, pp. 264–277, 2013.
[16] R. G. Kuehni, ‘‘How many object colors can we distinguish?’’ Color Res.
Appl., to be published, doi: 10.1002/col.21980.
[17] I. Sutherland and G. W. Hodgman, ‘‘Reentrant polygon clipping,’’
Commun. ACM, vol. 17, no. 1, pp. 32–42, 1974.
[18] B. Braden, ‘‘The surveyor’s area formula,’’ College Math. J., vol. 17, no. 4,
pp. 326–337, 1986.
[19] M. D. Fairchild, ‘‘Colorimetry,’’ in Color Appearance Models, 2nd ed.
Chichester, U.K.: Wiley, 2005, ch. 3, sec. 8, p. 80.
[20] K. Masaoka, ‘‘Single display gamut size metric,’’ J. Soc. Inf. Display, to
be published, doi: 10.1002/jsid.449.
[21] Information Display Measurements Standard, document IDMS Version
1.03, Society for Information Display, San Jose, CA, USA, 2012, p. 18.
[22] C. Poynton, ‘‘The CIE system of colorimetry,’’ in Digital Video and
HD, 2nd ed. Waltham, MA, USA: Morgan Kaufmann, 2012, ch. 25,
pp. 281–282.
[23] E. F. Kelley, K. Lang, L. D. Silverstein, and M. H. Brill, ‘‘A rational
methodology for estimating the luminous flux based upon color pri-
maries from digital projection displays,’’ NIST, Gaithersburg, MD, USA,
Tech. Rep. NISTIR 6657, 2009.
KENICHIRO MASAOKA received the B.S. degree
in electronics engineering and the M.S. degree
in energy engineering, and the Ph.D. degree in
engineering from the Tokyo Institute of Tech-
nology, Tokyo, Japan, in 1994 and 1996, and
2009, respectively. He joined Japan Broadcast-
ing Corporation (NHK), Tokyo, in 1996. He is
currently a Principal Research Engineer with the
Advanced Television Systems Research Division,
NHK Science and Technology Research Labora-
tories, Tokyo.
VOLUME 4, 2016