RGB coordinates of the Macbeth ColorChecker
Danny Pascale
The BabelColor Company
dpascale@BabelColor.com
www.BabelColor.com
© 2000-2006 Danny Pascale
Abstract. The ColorChecker chart, manufactured by GretagMacbeth,
is commonly used as a reference target for photographic and video
production work. This document provides RGB coordinates, in 8-bit
and 16-bit formats, for all color patches in four common RGB spaces
(Adobe, Apple, ProPhoto, and sRGB), which are defined in terms of
primaries, Illuminant, and gamma response. The method and equations
used to derive the data are presented as well. Reference data provided by
the chart’s manufacturer is compared to user-measured values.
Subject terms: ColorChecker, RGB coordinates, RGB space, color space,
color conversion.
Complete update
June 1st, 2006
RGB and ProPhoto, are not given. This paper’s purpose is to
provide numbers for these spaces, in both 8 bits per primary
(24 bits for R’G’B’) and 16 bits per primary formats (48 bits
for R’G’B’), as well as present the method by which they were
derived. These coordinates should be used in any program
where specific “RGB” values can be assigned. Please notice
the absence of primes against the letters of “RGB” in the
preceding sentence, which reflect how gamma corrected
coordinates are referred to in most software, even if R’G’B’ is
the correct form (albeit more cumbersome to write).
However, obtaining R’G’B’ data is not enough for many chart
users who want to know how representative these numbers
are, and how close these numbers are to the ones of their
chart. Also, inquisitive users may be interested in how the
new numbers compare with the old ones. This is the subject
of Section 2. The R’G’B’ values derived from the new
reference data are presented in Section 3, as well as the R’G’B’
values derived from the BabelColor average.
Section 4 presents a short description of each space. The
process by which the values were obtained follows, in
Section 5. The process can be used for spaces not covered by
this paper; for example, for a space defined for a particular
display with color primaries different from the ones presented
here.
2. Comparing ColorChecker references
We have gathered four data sets which we consider reliable
enough to be used as references:
i) ColorChecker 1976: xyY data with Illuminant C (Ref. 2)
ii) ColorChecker 2005: L*a*b* D50 and sRGB (Ref. 4)
iii) BabelColor Avg.: spectral data (Ref. 3)
iv) ProfileMaker 2004: spectral data (2/5/2004)
Other measurements were found, in either tristimulus or
spectral form, but they were either incomplete (no data for all
patches), or they were from a single chart, or their origin
could not be confirmed. We know that, some time ago,
L*a*b* D50 data was available from the Munsell Web site,
but the file was removed when the Web site was updated.
Introduction
1.
The ColorChecker1 chart is ubiquitous in the photographic
and video fields. Its main application is for obtaining a rapid
assessment of an imaging devices’ color rendering accuracy,
although it can also be used for simple calibration purposes.
The chart consists of 24 color patches formulated to emulate
common natural colors such as skin colors, foliage, and sky,
in addition to additive and subtractive primaries, and a six
steps gray scale2. While designed for optimum color
consistency when comparing pictures of the chart with
pictures of the natural colors, as reproduced on color film, it
was shown that the degree of metamerism was also very small
when directly comparing the chart to the natural colors.
Until recently, R’G’B’ values for this chart were difficult to
find. In particular, the R’G’B’ data supplied with the chart
corresponded to no common RGB space, and no primaries
and white point coordinates were provided either. The only
official and reliable tristimulus data supplied with the product
consisted of xyY coordinates measured with CIE Illuminant
C, a common Daylight Illuminant when the original data was
measured (from Ref. 2); this data was used, by this author, as
a basis to determine R’G’B’ values, and published in a
previous version of this document. In view of this limited
information, the author started, a few years ago, to compile
and average spectral data measured on ColorCheckers by
users from all over the world. This “real-life” data, from 20
charts of various ages (all of them in well kept conditions),
and measured using various instruments, can be seen as an
independent validation of the official reference data. This
average data, labeled “BabelColor Avg.” throughout this text,
was used to derive R’G’B’ values for comparison purposes.
Extracts of the data set are presented here; the complete data
is available in a spreadsheet which can be downloaded from
the BabelColor Web site3.
Since about October 2005, the GretagMacbeth Company
provides sRGB values, in 8-bit format, and L*a*b* D50 data
with its standard size and Mini format ColorChecker charts;
the data is also freely accessible on their Web site4. The
published data is the same for both charts. However, R’G’B’
data for other popular spaces, such as Adobe RGB, Apple
2
Table 1a
Color name
No.
dark skin
1
light skin
2
blue sky
3
foliage
4
blue flower
5
bluish green
6
orange
7
purplish blue
8
9 moderate red
10
11
12
13
14
15
16
17 magenta
18
19 white 9.5 (.05 D)
neutral 8 (.23 D)
20
neutral 6.5 (.44 D)
21
neutral 5 (.70 D)
22
neutral 3.5 (1.05 D)
23
black 2 (1.5 D)
24
purple
yellow green
orange yellow
blue
green
red
yellow
cyan
ColorChecker 2005
b*
L*
37,99
14,06
17,81
65,71
-21,93
49,93
21,91
43,14
-25,40
55,11
70,72
-0,199
57,10
62,66
-45,96
40,02
16,25
51,12
30,33
-21,59
57,26
72,53
67,86
71,94
-50,30
28,78
55,26
31,37
28,19
42,10
79,82
81,73
-14,57
51,94
51,04
-28,64
1,186
96,54
-0,335
81,26
-0,504
66,77
50,87
-0,270
-1,231
35,66
20,46
-0,973
a*
13,56
18,13
-4,88
-13,10
8,84
-33,40
36,07
10,41
48,24
22,98
-23,71
19,36
14,18
-38,34
53,38
4,04
49,99
-28,63
-0,425
-0,638
-0,734
-0,153
-0,421
-0,079
Table 1b
Color name
No.
dark skin
1
light skin
2
blue sky
3
foliage
4
blue flower
5
bluish green
6
orange
7
purplish blue
8
9 moderate red
10
11
12
13
14
15
16
17 magenta
18
19 white 9.5 (.05 D)
neutral 8 (.23 D)
20
neutral 6.5 (.44 D)
21
neutral 5 (.70 D)
22
neutral 3.5 (1.05 D)
23
black 2 (1.5 D)
24
purple
yellow green
orange yellow
blue
green
red
yellow
cyan
ColorChecker 2005
b*
L*
14,06
37,99
17,81
65,71
49,93
-21,93
21,91
43,14
-25,40
55,11
-0,199
70,72
62,66
57,10
-45,96
40,02
16,25
51,12
-21,59
30,33
57,26
72,53
71,94
67,86
-50,30
28,78
31,37
55,26
28,19
42,10
81,73
79,82
-14,57
51,94
-28,64
51,04
1,186
96,54
-0,335
81,26
66,77
-0,504
-0,270
50,87
-1,231
35,66
20,46
-0,973
a*
13,56
18,13
-4,88
-13,10
8,84
-33,40
36,07
10,41
48,24
22,98
-23,71
19,36
14,18
-38,34
53,38
4,04
49,99
-28,63
-0,425
-0,638
-0,734
-0,153
-0,421
-0,079
ColorChecker 1976
b*
L*
38,14
14,75
17,30
66,63
-22,43
50,73
21,85
43,36
-25,74
56,01
71,50
0,831
58,56
62,28
-44,07
40,44
15,56
51,94
30,50
-23,65
58,64
72,83
66,70
72,18
-52,83
28,59
55,66
33,09
26,98
41,71
78,47
81,95
-15,57
51,57
51,07
-27,36
0,067
96,00
0,058
81,35
0,049
66,67
51,58
0,040
0,031
35,99
0,022
20,56
avg. :
a*
13,81
15,38
-3,15
-14,99
9,63
-31,93
31,88
11,42
45,25
23,99
-23,76
17,40
20,31
-38,77
53,43
1,65
48,99
-28,01
-0,062
-0,054
-0,046
-0,037
-0,029
-0,020
L*
38,36
66,06
50,09
43,20
55,36
70,70
62,56
40,18
51,71
30,38
72,49
71,96
28,65
55,05
42,18
82,23
51,82
50,55
96,39
81,01
66,30
50,83
35,72
20,71
BabelColor Avg.
b*
14,65
17,85
-22,51
21,73
-24,82
-0,240
58,05
-44,29
16,86
-20,31
57,08
68,00
-50,52
31,62
28,79
79,84
-13,90
-28,14
2,238
0,180
-0,079
-0,268
-0,468
-0,447
avg. :
a*
13,80
17,74
-4,41
-13,46
8,89
-32,89
35,13
9,55
47,69
21,13
-23,46
19,49
15,60
-38,09
54,89
4,05
49,79
-27,97
-0,404
-0,570
-0,434
-0,687
-0,521
0,025
2006-06-01
ΔE
ΔE ≤ 1
1 < ΔE ≤ 2
2 < ΔE ≤ 4
4 < ΔE
ΔE
CIELAB DE2000
0,45
2,01
1,75
1,32
0,98
1,09
2,72
1,61
1,26
1,06
0,48
1,08
3,11
0,77
0,73
1,40
0,72
0,54
1,25
0,94
1,14
0,79
1,38
0,98
1,23
0,75
2,95
1,98
1,91
1,24
1,95
4,45
2,19
3,17
2,30
1,41
2,29
6,63
1,82
1,27
2,75
1,46
1,42
1,29
0,71
0,89
0,78
1,36
1,00
2,00
ΔE
CIELAB DE2000
0,74
0,52
0,77
0,41
0,63
0,51
1,34
1,89
1,01
2,24
0,31
0,19
1,44
0,41
1,63
0,50
0,71
0,96
1,06
0,57
0,70
0,54
0,77
0,59
0,85
0,49
0,40
0,50
0,31
0,46
0,21
0,80
0,39
0,73
0,97
0,11
0,07
0,88
0,27
0,45
0,34
0,32
0,57
0,98
0,54
0,71
0,78
0,75
0,56
0,52
Table 1a: Official L*a*b* D50 values of the ColorChecker, as made available by GretagMacbeth in 2005 (“ColorChecker 2005”),
compared to the previously distributed data measured in 1976 (“ColorChecker 1976”). The 1976 data, measured with
Illuminant C, was converted to Illuminant D50 using a Bradford chromatic adaptation transform.
Table 1b: “ColorChecker 2005” data compared to L*a*b* D50 data derived from the average of 20 charts
compiled by BabelColor (“BabelColor Avg.”).
2006-06-01
Table 1c
Color name
No.
dark skin
1
light skin
2
blue sky
3
foliage
4
blue flower
5
bluish green
6
orange
7
purplish blue
8
9 moderate red
10
11
12
13
14
15
16
17 magenta
18
19 white 9.5 (.05 D)
neutral 8 (.23 D)
20
neutral 6.5 (.44 D)
21
neutral 5 (.70 D)
22
neutral 3.5 (1.05 D)
23
black 2 (1.5 D)
24
purple
yellow green
orange yellow
blue
green
red
yellow
cyan
ColorChecker 2005
b*
L*
37,99
14,06
17,81
65,71
-21,93
49,93
21,91
43,14
-25,40
55,11
70,72
-0,199
57,10
62,66
-45,96
40,02
16,25
51,12
30,33
-21,59
57,26
72,53
67,86
71,94
-50,30
28,78
55,26
31,37
28,19
42,10
79,82
81,73
-14,57
51,94
51,04
-28,64
1,186
96,54
-0,335
81,26
-0,504
66,77
50,87
-0,270
-1,231
35,66
20,46
-0,973
a*
13,56
18,13
-4,88
-13,10
8,84
-33,40
36,07
10,41
48,24
22,98
-23,71
19,36
14,18
-38,34
53,38
4,04
49,99
-28,63
-0,425
-0,638
-0,734
-0,153
-0,421
-0,079
Table 1d
Color name
No.
dark skin
1
light skin
2
blue sky
3
foliage
4
blue flower
5
bluish green
6
orange
7
purplish blue
8
9 moderate red
10
11
12
13
14
15
16
17 magenta
18
19 white 9.5 (.05 D)
neutral 8 (.23 D)
20
neutral 6.5 (.44 D)
21
neutral 5 (.70 D)
22
neutral 3.5 (1.05 D)
23
black 2 (1.5 D)
24
purple
yellow green
orange yellow
blue
green
red
yellow
cyan
BabelColor Avg.
b*
14,65
17,85
-22,51
21,73
-24,82
-0,240
58,05
-44,29
16,86
-20,31
57,08
68,00
-50,52
31,62
28,79
79,84
-13,90
-28,14
2,238
0,180
-0,079
-0,268
-0,468
-0,447
a*
13,80
17,74
-4,41
-13,46
8,89
-32,89
35,13
9,55
47,69
21,13
-23,46
19,49
15,60
-38,09
54,89
4,05
49,79
-27,97
-0,404
-0,570
-0,434
-0,687
-0,521
0,025
L*
38,36
66,06
50,09
43,20
55,36
70,70
62,56
40,18
51,71
30,38
72,49
71,96
28,65
55,05
42,18
82,23
51,82
50,55
96,39
81,01
66,30
50,83
35,72
20,71
ProfileMaker 2004
b*
L*
38,40
14,52
18,22
66,07
-21,70
50,17
22,67
43,27
-25,14
55,47
71,23
-0,060
58,29
62,83
-45,87
40,27
16,56
51,26
30,47
-21,28
57,89
72,95
68,94
72,27
-50,02
28,71
55,40
32,01
28,41
41,50
80,85
82,56
-14,30
52,20
51,37
-28,29
1,470
96,96
0,097
81,57
-0,083
67,17
50,15
-0,490
-0,741
35,94
-0,365
20,38
avg. :
a*
13,58
18,02
-4,91
-13,33
8,84
-33,03
35,88
10,39
48,01
21,07
-23,45
19,25
14,36
-38,02
56,42
3,49
49,90
-28,48
-0,474
-0,703
-0,779
-1,225
-0,336
-0,360
ProfileMaker 2004
b*
L*
14,52
38,40
18,22
66,07
50,17
-21,70
22,67
43,27
-25,14
55,47
-0,060
71,23
62,83
58,29
-45,87
40,27
16,56
51,26
-21,28
30,47
57,89
72,95
72,27
68,94
-50,02
28,71
32,01
55,40
28,41
41,50
82,56
80,85
-14,30
52,20
-28,29
51,37
1,470
96,96
0,097
81,57
67,17
-0,083
-0,490
50,15
-0,741
35,94
-0,365
20,38
avg. :
a*
13,58
18,02
-4,91
-13,33
8,84
-33,03
35,88
10,39
48,01
21,07
-23,45
19,25
14,36
-38,02
56,42
3,49
49,90
-28,48
-0,474
-0,703
-0,779
-1,225
-0,336
-0,360
3
ΔE
ΔE ≤ 1
1 < ΔE ≤ 2
2 < ΔE ≤ 4
4 < ΔE
ΔE
CIELAB DE2000
0,47
0,43
0,27
0,40
0,38
0,43
0,55
0,23
0,26
1,03
0,43
0,42
0,25
0,37
1,14
0,69
0,29
0,36
0,37
0,49
0,53
1,71
0,54
0,73
0,53
0,62
0,56
0,33
0,81
0,45
0,65
1,22
0,27
0,41
1,94
0,80
1,14
0,34
0,73
3,11
1,43
0,39
0,51
0,51
0,54
0,58
1,31
0,58
0,68
0,83
ΔE
CIELAB DE2000
0,26
0,47
0,95
0,95
0,34
0,58
0,83
1,79
0,62
0,98
0,93
1,02
1,33
0,53
1,72
1,20
0,56
0,97
0,96
0,58
0,94
0,89
0,40
0,51
0,85
0,17
0,23
0,55
0,55
0,26
0,43
0,41
0,35
0,50
0,61
0,43
0,43
0,65
0,38
0,90
0,47
0,41
0,85
0,80
0,44
0,86
1,04
0,42
0,62
0,53
Table 1c: Official L*a*b* D50 values of the ColorChecker, as made available by GretagMacbeth in 2005 (“ColorChecker 2005”),
compared to values derived from the reference spectral file provided with ProfileMaker (“ProfileMaker 2004”;
Table 1d: L*a*b* D50 data derived from the average of 20 charts compiled by BabelColor (“BabelColor Avg.”),
file name: “ColorChecker 24”; file measurement date: “2/5/2004”).
compared to “ProfileMaker 2004”.
4
2006-06-01
The first comparison that comes to mind is the one between
the two “official” tristimulus data sets, published in 1976
(“ColorChecker 1976”) and 2005 (“ColorChecker 2005”),
shown in Table 1a. In order to compare the two sets on the
same basis, we have chosen to convert the xyY Ill-C (1976)
coordinates to the L*a*b* D50 color space used for the most
recent reference. From xyY, one can readily determine XYZ
values, then use a Chromatic Adaptation Transform (CAT), in
this case the Bradford matrix discussed in Section 5.2, to
convert the XYZ coordinates between Illuminant C and
Illuminant D50, and finally compute the proper L*a*b*
values. The use of a CAT is required since we do not have the
spectral data corresponding to the xyY coordinates. While
using a CAT can introduce an error, this error has less of an
effect than if it was simply added to the inherent difference
between the data sets; see Section 6 for more information.
third comparison, Table 1c,
The second comparison, shown in Table 1b, is between the
“ColorChecker 2005” data set and the “BabelColor Avg.”.
The
the
“ColorChecker 2005” set and tristimulus data derived from a
spectral reference file of the ColorChecker (ProfileMaker
2004). This file is provided by GretagMacbeth as part of their
ProfileMaker software package; the measurement date shown
in the file is “2/5/2004”. The fourth comparison, Table 1d, is
between the “BabelColor Avg.” and “ProfileMaker 2004”
data sets.
is between
The color differences in Table 1 are computed using both
CIELAB and CIEDE2000. CIEDE2000 is the most recent
color difference formula recommended by the Commission
Internationale de l'Éclairage (CIE). Like the CIE94 and CMC
color difference formulas which came after CIELAB, it
strives to improve the match between the perceived color
difference and the computed difference values. CIEDE2000,
similarly to the CIE94 and CMC formulas, includes weighting
functions for
it
introduces an extra term which combines chroma and hue
with the goal of improving the performance for blue colors
(for hue angles – the h* in the L*C*h* presentation format –
around 275 degrees). It also associates a scaling factor to a*
for low chroma colors, to improve the formula performance
near the
illuminant. Many users have confirmed that
CIEDE2000, while still not perfect, does achieve its goal of
improving the match between computed difference numbers
and perceived difference5.
lightness, chroma and hue. However,
In Table 1a, we see a noticeable difference between the
“ColorChecker 1976” and “ColorChecker 2005” data sets,
whereas the difference is quite small when comparing the
2005 data with either the “BabelColor Avg.” or the
“ProfileMaker 2004” sets in Tables 1b and 1c. The 1976 data
may have been deemed sufficiently precise at a time where the
chart was mostly used to visually judge the quality of silver-
based
to make precise digital
measurements as we do now.
films, and not used
As per GretagMacbeth Web site, the 2005 ColorChecker data
“is intended to be an average measurement of all ColorChecker Charts”.
The fact that, on average, this data set cannot be visually
differentiated from either the “ProfileMaker 2004” or the
“BabelColor Avg.” data sets makes it difficult to select the
best one. There is no detailed information on where the 2005
data comes from; it may be an average from one, or from
many production lots. There is even less information on the
origin of the ProfileMaker reference file but its good match to
the other data sets indicates it is also an average of some sorts.
As for the data compiled by BabelColor, the match to the
other two data sets is quite good, especially considering the
mix of experimental conditions imposed by many users using
different instruments. Overall, the similarity of the three data
sets points to some outstanding long term production
consistency.
Readers interested in seeing spectral graphs for each patch, as
well as information on spectral and L*a*b* variance, can
download
“ColorChecker_RGB_and_spectra.xls”
spreadsheet from the BabelColor Web site (see Ref. 3).
the
3. RGB coordinates of the ColorChecker
The R’G’B’ values of the ColorChecker for four common
RGB spaces, Adobe, Apple, ProPhoto and sRGB, are shown
in 8-bit format in Table 2, and in 16-bit format in Table 3.
Table 3 is a more precise version of Table 2, with more
significant digits per value. The 16-bit values can be used
mainly in programming environments, such as MATLAB,
since there is no color picker that yet offers 16-bit resolution.
You should be aware that, for computing efficiency reasons,
Photoshop processes 16-bit file as if 15-bit and resaves the
file as 16-bit; the displayed color numbers are thus divided by
two from the 16-bit values.
In Tables 2 and 3, the tables labeled “ColorChecker 2005”
show the L*a*b* D50 values provided by GretagMacbeth.
You will notice two columns with sRGB in their title in
Table 2; the one labeled “sRGB (GMB)” contains the values
provided by GretagMacbeth, while the “sRGB” column was
derived from L*a*b* D50 using the procedure presented in
Section 5. The other R’G’B’ values of the “ColorChecker
2005” table were derived in a similar manner. It should be
emphasized that for ProPhoto, a D50 based RGB space, there
is no need to perform a chromatic adaptation transform when
starting with L*a*b* D50 and that there
is minimal
“conversion process-induced” errors (see Section 6).
All R’G’B’ values of the “BabelColor Avg.” tables were
obtained with the spectral reflectance average of 20 charts, the
space Illuminant spectral distribution, and the 2-degrees
Standard Observer. In other words, they were not obtained
using a chromatic adaptation transform, and do not comprise
the errors this transform may introduce.
It is interesting to note in Table 2 that the “sRGB (GMB)”
cyan patch is measured to be within the sRGB gamut, with an
R’ value of 8, while this coordinate is clipped to zero when
derived from the L*a*b* data (as can be seen in the “sRGB”
column of the “ColorChecker 2005” table). The cyan is
similarly clipped in the “BabelColor Avg.” tables.
2006-06-01
ColorChecker 2005
No.
Color name
illuminant
0
dark skin
1
light skin
2
blue sky
3
foliage
4
blue flower
5
bluish green
6
orange
7
purplish blue
8
9 moderate red
purple
10
yellow green
11
orange yellow
12
blue
13
green
14
red
15
yellow
16
17 magenta
cyan
18
19 white 9.5 (.05 D)
neutral 8 (.23 D)
20
neutral 6.5 (.44 D)
21
neutral 5 (.70 D)
22
neutral 3.5 (1.05 D)
23
black 2 (1.5 D)
24
BabelColor Avg.
Color name
No.
illuminant
0
dark skin
1
light skin
2
blue sky
3
foliage
4
blue flower
5
bluish green
6
orange
7
purplish blue
8
9 moderate red
purple
10
yellow green
11
orange yellow
12
blue
13
green
14
red
15
yellow
16
17 magenta
cyan
18
19 white 9.5 (.05 D)
neutral 8 (.23 D)
20
neutral 6.5 (.44 D)
21
neutral 5 (.70 D)
22
neutral 3.5 (1.05 D)
23
black 2 (1.5 D)
24
5
Apple
sRGB
ProPhoto
Adobe
sRGB (GMB)
R' G' B' R' G' B' R' G' B' R' G' B' R' G' B'
255
255
68
107
130
184
157
101
67
95
128
177
170
129
44
201
166
77
174
99
108
86
64
167
46
213
150
49
99
73
60
155
31
227
149
169
161
61
245
242
200
200
160
160
121
120
84
85
52
52
255
70
129
153
69
173
171
56
166
97
104
75
55
143
80
59
52
147
167
242
201
162
121
86
54
255
82
150
122
108
128
189
126
91
90
60
188
163
61
148
54
199
86
133
243
200
160
122
85
52
255
115
194
98
87
133
103
214
80
193
94
157
224
56
70
175
231
187
8
243
200
160
122
85
52
255
67
129
156
64
176
172
47
170
97
106
63
39
147
74
57
20
151
170
243
202
163
122
86
51
255
81
147
122
108
128
190
124
91
82
58
189
162
63
149
49
198
84
136
245
202
163
121
84
49
255
116
199
91
90
130
92
224
68
198
94
159
230
35
67
180
238
193
0
245
200
161
121
82
49
255
54
114
133
55
154
157
54
145
80
82
74
60
120
69
46
66
127
148
240
191
146
102
68
38
255
67
135
102
86
111
168
118
74
83
49
170
152
50
123
59
188
85
111
243
190
146
102
66
37
255
81
159
94
75
118
127
167
79
141
68
144
181
57
85
120
199
143
78
242
189
145
102
66
37
255
51
109
139
48
162
155
30
156
79
88
39
19
131
54
43
0
134
154
239
191
146
102
68
38
255
63
128
103
89
108
178
102
71
59
42
177
143
47
133
29
187
60
118
243
191
146
102
66
37
255
94
183
74
73
110
84
211
52
180
73
145
220
26
60
159
232
174
0
242
189
144
101
65
37
255
82
146
122
107
127
188
123
92
83
61
188
160
65
148
52
197
85
135
245
201
161
120
85
53
Apple
ProPhoto
sRGB
Adobe
255
70
129
155
69
173
170
54
164
97
103
75
56
143
78
59
53
147
164
240
199
160
121
85
54
R' G' B' R' G' B' R' G' B' R' G' B'
255
255
67
107
182
129
157
103
65
96
176
129
171
132
197
45
168
79
98
171
105
84
62
168
211
41
147
53
72
101
56
151
22
227
150
165
166
65
240
245
201
199
159
161
121
119
85
84
53
50
255
81
149
123
108
129
191
123
92
84
59
189
161
62
149
48
200
84
136
245
201
161
121
84
50
255
115
196
93
90
130
99
220
72
195
91
160
229
43
71
176
238
188
0
245
200
160
120
83
50
255
54
114
135
56
154
157
52
143
80
81
74
60
120
68
46
67
126
146
237
189
144
102
67
38
255
68
136
102
86
111
168
118
75
85
50
169
152
50
122
59
190
85
109
242
190
144
102
67
37
255
82
160
95
75
119
127
166
79
142
67
144
182
57
85
121
201
143
77
242
189
144
101
66
37
255
51
109
141
48
162
154
29
153
79
87
38
20
131
52
43
0
133
151
236
189
144
102
67
38
255
63
130
103
90
108
179
101
72
62
43
177
142
45
133
29
188
60
118
242
190
144
102
67
37
255
94
180
76
73
110
89
206
55
176
70
147
218
32
63
154
232
168
0
243
188
143
101
66
37
255
82
148
122
108
128
189
122
92
85
61
188
159
64
148
52
198
85
135
245
200
160
121
85
53
xyY (CIE D50)
x
y
xyY (CIE D50)
x
y
Y
100
L*a*b* (CIE D50)
b*
L*
100
0
0,3585
14,06
0,3777 10,08 37,99
0,3744 34,95 65,71
17,81
-21,93
0,3016 18,36 49,93
0,4499 13,25 43,14
21,91
-25,40
0,2856 23,04 55,11
-0,20
0,3911 41,78 70,72
0,4055 31,18 62,66
57,10
-45,96
0,2106 11,26 40,02
0,3273 19,38 51,12
16,25
-21,59
0,2482
6,37 30,33
57,26
0,5008 44,46 72,53
0,4427 43,57 71,94
67,86
-50,30
0,1692
5,75 28,78
0,5032 23,18 55,26
31,37
28,19
0,3303 12,57 42,10
0,4734 59,81 81,73
79,82
-14,57
0,2688 20,09 51,94
-28,64
0,3023 19,30 51,04
0,3608 91,31 96,54
1,19
-0,34
0,3584 58,94 81,26
-0,50
0,3581 36,32 66,77
-0,27
0,3579 19,15 50,87
0,3548
8,83 35,66
-1,23
-0,97
3,11 20,46
0,3537
a*
0
13,56
18,13
-4,88
-13,10
8,84
-33,40
36,07
10,41
48,24
22,98
-23,71
19,36
14,18
-38,34
53,38
4,04
49,99
-28,63
-0,43
-0,64
-0,73
-0,15
-0,42
-0,08
Y
100
L*a*b* (CIE D50)
b*
L*
0
0,3585
100
14,65
0,3787 10,29 38,36
0,3749 35,40 66,06
17,85
-22,51
0,2996 18,49 50,09
0,4501 13,29 43,20
21,73
-24,82
0,2871 23,28 55,36
-0,24
0,3905 41,75 70,70
0,4081 31,06 62,56
58,05
-44,29
0,2157 11,36 40,18
0,3295 19,89 51,71
16,86
-20,31
0,2544
6,39 30,38
57,08
0,5002 44,40 72,49
0,4426 43,60 71,96
68,00
-50,52
0,1676
5,70 28,65
31,62
0,5040 22,97 55,05
28,79
0,3284 12,62 42,18
0,4730 60,72 82,23
79,84
-13,90
0,2702 19,98 51,82
-28,14
0,3028 18,89 50,55
2,24
0,3625 90,94 96,39
0,18
0,3593 58,50 81,01
0,3588 35,71 66,30
-0,08
-0,27
0,3586 19,12 50,83
-0,47
8,87 35,72
0,3577
0,3562
3,17 20,71
-0,45
a*
0
13,80
17,74
-4,41
-13,46
8,89
-32,89
35,13
9,55
47,69
21,13
-23,46
19,49
15,60
-38,09
54,89
4,05
49,79
-27,97
-0,40
-0,57
-0,43
-0,69
-0,52
0,03
0,3457
0,4316
0,4197
0,2760
0,3703
0,2999
0,2848
0,5295
0,2305
0,5012
0,3319
0,3984
0,4957
0,2018
0,3253
0,5686
0,4697
0,4159
0,2131
0,3469
0,3440
0,3432
0,3446
0,3401
0,3406
0,3457
0,4336
0,4187
0,2757
0,3688
0,3016
0,2856
0,5291
0,2335
0,5002
0,3316
0,3986
0,4960
0,2042
0,3262
0,5734
0,4693
0,4175
0,2146
0,3486
0,3451
0,3447
0,3433
0,3425
0,3436
Table 2: R’G’B’ coordinates of the ColorChecker, in 8-bit format. Coordinates for which clipping occurred are shown with a gray background.
Top (“ColorChecker 2005”): The L*a*b* and “sRGB (GMB)” data is from GretagMacbeth; the other values (xyY, Adobe,
Apple, ProPhoto and sRGB) were derived from the L*a*b* data using the procedure described in Section 5.
Bottom (“BabelColor Avg.”): L*a*b* and R’G’B’ values were derived from the spectral average of 20 charts.
6
ColorChecker 2005
No.
Color name
illuminant
0
dark skin
1
light skin
2
blue sky
3
foliage
4
blue flower
5
bluish green
6
orange
7
purplish blue
8
9 moderate red
10
11
12
13
14
15
16
17 magenta
18
19 white 9.5 (.05 D)
neutral 8 (.23 D)
20
neutral 6.5 (.44 D)
21
neutral 5 (.70 D)
22
neutral 3.5 (1.05 D)
23
black 2 (1.5 D)
24
purple
yellow green
orange yellow
blue
green
red
yellow
cyan
BabelColor Avg.
Color name
No.
illuminant
0
dark skin
1
light skin
2
blue sky
3
foliage
4
blue flower
5
bluish green
6
orange
7
purplish blue
8
9 moderate red
10
11
12
13
14
15
16
17 magenta
18
19 white 9.5 (.05 D)
neutral 8 (.23 D)
20
neutral 6.5 (.44 D)
21
neutral 5 (.70 D)
22
neutral 3.5 (1.05 D)
23
black 2 (1.5 D)
24
purple
yellow green
orange yellow
blue
green
red
yellow
cyan
a*
0
L*a*b* (CIE D50)
b*
L*
100
0
37,986
65,711
49,927
43,139
55,112
70,719
62,661
40,020
51,124
30,325
72,532
71,941
28,778
55,261
42,101
81,733
51,935
51,038
96,539
81,257
66,766
50,867
35,656
20,461
13,555
18,130
-4,880
-13,095
8,844
-33,397
36,067
10,410
48,239
22,976
-23,709
19,363
14,179
-38,342
53,378
4,039
49,986
-28,631
-0,425
-0,638
-0,734
-0,153
-0,421
-0,079
14,059
17,810
-21,925
21,905
-25,399
-0,199
57,096
-45,964
16,248
-21,587
57,255
67,857
-50,297
31,370
28,190
79,819
-14,574
-28,638
1,186
-0,335
-0,504
-0,270
-1,231
-0,973
a*
0
L*a*b* (CIE D50)
b*
L*
100
0
38,358
66,056
50,090
43,204
55,356
70,700
62,559
40,178
51,711
30,375
72,492
71,963
28,653
55,046
42,182
82,230
51,820
50,555
96,387
81,014
66,297
50,830
35,724
20,706
13,802
17,737
-4,407
-13,464
8,891
-32,892
35,135
9,551
47,694
21,131
-23,462
19,486
15,600
-38,088
54,893
4,048
49,787
-27,973
-0,404
-0,570
-0,434
-0,687
-0,521
0,025
14,646
17,848
-22,512
21,730
-24,824
-0,240
58,050
-44,289
16,857
-20,309
57,078
67,998
-50,520
31,617
28,785
79,844
-13,904
-28,139
2,238
0,180
-0,079
-0,268
-0,468
-0,447
Adobe
G'
65535
21037
37489
31234
27576
32643
48441
31575
23542
21283
15563
48199
41157
16824
37925
13377
50657
21777
34706
62936
51637
41470
30941
21809
13539
Adobe
G'
65535
21170
37930
31323
27679
32793
48634
31443
23623
21847
15754
48202
40883
16411
37972
13259
50936
21730
34708
62839
51454
41102
30992
21851
13647
B'
65535
17932
33025
39364
17662
44447
44028
14313
42751
24831
26707
19356
14181
36877
20582
15245
13301
37827
42920
62298
51689
41576
31029
22201
13841
B'
65535
17940
33063
39736
17760
44393
43737
13962
42180
24990
26485
19192
14408
36835
20042
15090
13593
37690
42055
61577
51238
41054
30991
21954
13795
R'
65535
27426
47379
25919
24528
32927
33187
51626
19711
44624
22093
42816
54654
12591
25519
39846
58361
43542
15780
62890
51294
41082
30850
21522
13424
R'
65535
27427
46858
26357
24611
33111
34033
50616
20328
43878
21538
43111
54237
13494
26024
38918
58390
42397
16727
62871
51214
40921
30704
21631
13618
R'
65535
24272
47061
19061
18814
28179
21635
54315
13412
46277
18775
37223
56662
6744
15499
40882
59657
44678
0
62308
48454
36900
26064
16716
9404
R'
65535
24245
46219
19634
18888
28387
22962
52933
14223
45214
18091
37668
56124
8096
16294
39705
59620
43169
0
62323
48400
36794
25852
16842
9595
Table 3: R’G’B’ coordinates of the ColorChecker, in 16-bit format. Coordinates for which clipping occurred are shown with a gray background.
Top (“ColorChecker 2005”): The L*a*b* data is from GretagMacbeth; the R’G’B’values were derived from the L*a*b* data
Bottom (“BabelColor Avg.”): L*a*b* and R’G’B’ values were derived from the spectral average of 20 charts.
using the procedure described in Section 5.
Apple
ProPhoto
sRGB
2006-06-01
G'
65535
16153
32784
26484
22981
27684
45841
26099
18365
15231
10808
45581
36855
11979
34177
7492
47961
15483
30411
62388
48985
37472
26190
17081
9535
G'
65535
16287
33312
26559
23087
27846
46056
26017
18452
15889
11025
45576
36556
11556
34223
7483
48296
15524
30420
62281
48775
37062
26250
17124
9628
B'
65535
13188
27946
35753
12230
41629
39917
7836
40089
20218
22518
10085
4761
33587
13894
11082
0
34453
39705
61541
49038
37587
26286
17484
9817
B'
65535
13185
27949
36178
12320
41556
39544
7462
39419
20351
22277
9858
5255
33552
13271
10945
0
34301
38682
60612
48492
36998
26240
17226
9768
R'
65535
20795
40907
24175
19249
30315
32629
42868
20351
36263
17495
37022
46640
14563
21842
30812
51240
36857
19993
62217
48688
37161
26131
16932
9502
G'
65535
17235
34660
26096
22117
28442
43161
30308
19117
21340
12667
43576
38989
12920
31642
15251
48436
21856
28496
62346
48923
37407
26179
17053
9531
ProPhoto
R'
65535
21065
41088
24358
19220
30507
32731
42582
20337
36595
17247
37047
46688
14720
21764
31163
51670
36739
19852
62107
48512
36885
26002
16963
9628
G'
65535
17418
34953
26172
22181
28596
43109
30346
19254
21774
12835
43524
38992
12758
31474
15087
48849
21803
28125
62194
48708
37024
26198
17093
9630
B'
65535
13942
29175
34288
14249
39678
40445
13797
37271
20465
21146
19080
15365
30946
17800
11935
16899
32665
38002
61675
49018
37556
26268
17412
9770
B'
65535
13958
29396
34655
14334
39609
40449
13468
36662
20603
20734
19115
15334
30949
17600
11809
17164
32307
37418
60977
48565
37019
26243
17205
9743
R'
65535
29684
51033
23285
23061
33299
23760
57637
17444
50970
24062
40800
59221
9090
17200
46236
61244
49611
0
62954
51492
41301
31014
21187
12507
R'
65535
29648
50244
23958
23139
33501
25382
56443
18426
50002
23277
41244
58773
11084
18307
45152
61195
48247
0
62967
51453
41222
30782
21331
12759
G'
65535
20794
37831
31447
27664
32893
48805
31797
23445
21055
14904
48564
41533
16275
38272
12506
50998
21580
35002
63018
51965
41847
31145
21613
12685
sRGB
G'
65535
20935
38278
31538
27772
33047
48996
31662
23530
21654
15113
48567
41257
15826
38320
12376
51274
21529
35005
62924
51785
41477
31198
21658
12804
B'
65535
17311
33071
40035
16548
45254
44209
12000
43738
24945
27134
16148
10089
37805
19051
14638
5069
38695
43613
62371
52019
41958
31239
22046
13032
B'
65535
17313
33089
40419
16655
45194
43894
11558
43150
25096
26893
15925
10499
37768
18395
14467
5688
38553
42717
61645
51563
41427
31196
21772
12975
Apple
2006-06-01
7
sRGB from L*a*b* D50
(ColorChecker 2005)
sRGB (GMB)
(ColorChecker 2005)
No.
Color name
R' G' B'
dark skin
1
light skin
2
blue sky
3
foliage
4
blue flower
5
bluish green
6
orange
7
purplish blue
8
9 moderate red
10
11
12
13
14
15
16
17 magenta
18
19 white 9.5 (.05 D)
neutral 8 (.23 D)
20
neutral 6.5 (.44 D)
21
neutral 5 (.70 D)
22
neutral 3.5 (1.05 D)
23
black 2 (1.5 D)
24
purple
yellow green
orange yellow
blue
green
red
yellow
cyan
81
116
67
199 147 129
122 156
91
90
108
64
130 128 176
190 172
92
47
224 124
68
91
170
97
82
198
106
94
58
63
159 189
39
230 162
35
63
147
74
149
67
57
180
49
20
238 198
193
84
151
0
136 170
245 245 243
200 202 202
161 163 163
121 121 122
86
82
51
49
84
49
R' G' B'
82
98
87
L*a*b* (CIE D65)
b*
L*
14,07 115
37,85
68
17,21 194 150 130
65,43
122 157
-21,79
50,15
43,17
22,44
108
67
-25,06 133 128 177
55,40
0,29 103 189 170
70,92
44
56,24 214 126
62,06
80
40,59
-45,14
91
166
99
90
15,17 193
50,58
108
-21,74
30,51
94
60
64
57,83 157 188
72,31
46
67,80 224 163
71,43
56
29,46
-49,34
61
150
73
148
32,03
55,26
70
60
26,92 175
41,53
54
31
80,10 231 199
81,08
86
-15,48 187
51,74
149
52,41
-26,64
8
133 161
0,96 243 243 242
96,49
-0,24 200 200 200
81,17
-0,25 160 160 160
66,84
50,86
-0,55 122 122 121
85
-1,44
35,61
52
20,40
-1,27
a*
12,72
17,18
-1,91
-15,08
11,58
-33,22
33,37
16,15
47,55
25,11
-27,84
15,50
20,74
-41,23
52,67
-0,33
51,26
-18,46
-0,35
-0,69
-0,71
0,20
-0,36
0,47
85
52
85
52
L*a*b* (CIE D65)
b*
L*
13,66
38,02
16,90
65,67
-21,60
50,63
43,00
20,45
-25,17
55,68
1,54
70,99
56,13
61,14
41,12
-41,88
14,88
51,33
-22,10
31,10
56,96
71,90
64,91
71,04
30,35
-49,67
32,30
55,03
24,66
41,35
77,55
80,70
-15,28
51,14
51,15
-23,37
0,48
95,82
0,00
80,60
0,00
65,87
51,19
0,55
0,00
36,15
0,00
21,70
avg. :
a*
11,80
13,67
0,37
-15,88
12,76
-30,64
28,10
17,41
42,10
24,35
-28,10
12,60
26,43
-40,14
49,30
-3,66
48,15
-19,73
-0,18
0,00
0,00
-0,20
0,00
0,00
CIELAB
ΔE*ab
1,02
3,52
2,34
2,15
1,23
2,87
5,35
3,53
5,52
1,02
1,01
4,11
5,77
1,15
4,06
4,22
3,17
3,72
0,84
0,92
1,23
1,21
1,58
1,89
2,64
Table 4: Color difference between the sRGB coordinates derived from L*a*b* D50 values provided by GretagMacbeth, and the
sRGB coordinates also provided by GretagMacbeth. The R’G’B’ coordinates of the “sRGB from L*a*b* D50” data set were
rounded to the nearest integer before computing the color differences. The L*a*b* values are computed for D65.
Because the cyan patch is close to the edge of the space
gamut, we could expect to have some measurements which
cross the border now and then; however, we have verified
that the 20 individual R’G’B’ values of the cyan patches used
for the “BabelColor Avg.” were all clipped. As an added
check, for all ColorChecker patches, we compared the sRGB
1st R'G'B'
data set
2nd R'G'B'
data set
Illum. for
ΔE*ab
avg.
ΔE*ab
sRGB from L*a*b* D50
(ColorChecker 2005)
sRGB (GMB)
(ColorChecker 2005)
D65
2,64
1
2
3
sRGB from L*a*b* D50
(ColorChecker 2005)
sRGB
(BabelColor Avg.)
sRGB (GMB)
(ColorChecker 2005)
sRGB
(BabelColor Avg.)
4 Adobe RGB from L*a*b* D50
(ColorChecker 2005)
Adobe RGB
(BabelColor Avg.)
5 Apple RGB from L*a*b* D50
(ColorChecker 2005)
Apple RGB
(BabelColor Avg.)
6
ProPhoto from L*a*b* D50
(ColorChecker 2005)
ProPhoto
(BabelColor Avg.)
D65
1,30
D65
1,95
D65
1,30
D65
1,25
D50
1,03
Table 5: Average CIELAB color differences of the 24 patches
of the ColorChecker for various R’G’B’ data sets. The color
difference is computed for the illuminant of the RGB space.
See Table 4 for the details of how the result of the first row,
“sRGB from L*a*b* D50” vs “sRGB (GMB)”, was obtained.
coordinates derived from L*a*b* D50 to the values provided
by GretagMacbeth (i.e. “sRGB (GMB)”). The individual color
differences and their average are shown in Table 4. Many
individual differences are large, with a maximum of 5,77 for
the blue patch; this is quite high, even for a blue CIELAB
difference. We have further compared the differences
between the other R’G’B’ data sets of Table 2; the averages
are shown in Table 5. For the sRGB space, the smallest
average color difference (=1,30) is seen between the “sRGB
from L*a*b* D50 (ColorChecker 2005)” and the “sRGB
(BabelColor Avg.)” data sets.
As we mentioned, the “sRGB from L*a*b* D50” data of the
“ColorChecker 2005” tables was determined using the
Bradford chromatic adaptation transform. In comparison, the
“sRGB (GMB)” data was likely determined from the spectral
reflectance data, the D65 Illuminant spectral distribution and
the 2-degrees Standard Observer, a method which is generally
more precise. A small numeric difference is thus expected
between the two methods. However, the actual average
difference between these two data sets is too high (=2,64) to
be explained only by chromatic transform errors only. A
rough estimate of the error introduced by the Bradford
transform can be obtained by comparing the average errors of
the D50 and D65 spaces in Table 5, since all D65 data derived
“from L*a*b* D50” was done so using the Bradford
transform. When comparing the sets of rows #2, 4 and 5 to
the sets of row #6, the D65 averages are 0,22 (=1,25-1,03)
8
2006-06-01
and 0,27 (=1,30-1,03) higher than the D50 data sets (row #6).
This small difference is due essentially to the Bradford
transform applied to L*a*b* D50 data. From this, we infer
that the sRGB values provided by GretagMacbeth came from
another data set than the one used for their L*a*b* values.
Overall, it can be seen that there is excellent agreement
between the R’G’B’ values of the “ColorChecker 2005” data
sets derived from L*a*b* values and the “BabelColor Avg.”
data sets (rows #2, 4, 5, and 6). This is in fact just another
way to look at what was shown in Table 1b. The best match
is, as expected, between the ProPhoto data sets, since no
chromatic adaptation transform was required in processing
the “ColorChecker 2005” L*a*b* D50 data.
It is important to note that all these differences between data
sets do not indicate which set is the best. However, the better
match between the “sRGB from L*a*b* D50 (ColorChecker
2005)” and the “sRGB (BabelColor Avg.)” data sets, when
compared to the large difference between the two sRGB data
sets of the “ColorChecker 2005” table, tend to indicate that
the “sRGB (GMB)” values are less reliable.
R’G’B’ values for many other common and uncommon
spaces can be found in Ref. 3.
4. RGB spaces descriptions
RGB spaces have evolved, sometimes for technological
reasons (NTSC evolved to SMPTE-C), sometimes to fulfill
professional requirements (ColorMatch, Adobe RGB), and
sometimes because that’s how the display system was built
and it became a, de-facto, standard (Apple RGB).
A short description of the four spaces selected for Tables 2
and 3 follows. The position of their primaries on a CIE 1931
chromaticity diagram can be seen in Figure 1. Numerical
specifications for each space are shown in Table 6.
Adobe RGB (1998)
Formerly known as SMPTE-240M for Photoshop user, this
space has been renamed once the final SMPTE-240M
standard committee settled for a smaller gamut6. Adobe RGB
is very close to the original NTSC space and has a large
enough gamut that encompasses the gamuts of most printing
processes and displays. However, because of its gamut size,
16 bits per primary file formats should be preferred to 8 bits
per primary ones, especially for editing purposes. While a
relatively large number of those colors cannot be printed
using the SWOP process (SWOP: Specifications for Web
Offset Publications), particularly in the green portion of the
gamut, newer printing processes,
as Pantone
Hexachrome, take advantage of this space. Adobe RGB’s
white is defined with Illuminant D65.
such
Apple RGB
Once a very common RGB space on the desktop, it is now
slowly getting phased out and replaced by sRGB, for everyday
use, and by Adobe RGB (and other larger gamut spaces) for
photographic and graphic design applications. Its gamut size
is similar to the ones of the ColorMatch and sRGB spaces.
lookup-table (LUT) gamma which
The Apple RGB, like the ColorMatch and SGI spaces, has a
non-unity display
is
compensated by the file encoding gamma (see Section 5.4 for
a discussion of gamma). In older Macintosh computers, when
a value of 1,8 was entered by the user in the control panel for
display gamma,
filled with numbers
corresponding to a gamma equal to 1,8/2,6=0,69 (or 1,45 if
you define gamma using the reciprocal value =1/0,69).
the LUT was
ColorSync, Apple’s color management technology at the
operating system level, now takes care of color management
for all input and output devices and will automatically convert
color data from one space to another for compliant
applications. Apple RGB’s white is Illuminant D65.
ProPhoto
ProPhoto is a very large gamut RGB space designed by
Kodak; it is getting attention from digital camera users as an
archiving and working space for RAW (minimally processed,
high dynamic range, and un-color-balanced) camera data.
Formerly called ROMM RGB while being developed, it was
renamed at the same time as its gamma was changed from 2,2
(=1/0,455) to 1,8 (=1/0,556). ProPhoto’s white is Illuminant
D50.
While it covers most of the visible spectrum, it also extends
outside of it. As a result, about 13% of the RGB triads
represent non-existent colors. Working at 16 bits per channel
is a minimum with this space, and some users are concerned
that even this bit depth is not enough. Others are puzzled by
the decision to use a 1,8 gamma when the industry is slowly
moving towards a standard 2,2 value. In any case, when used
with caution for images that DO contain colors outside of the
range of medium size working spaces, like Adobe RGB, it can
provide improved color rendering when used in conjunction
with modern wide gamut inkjet printers.
sRGB
With chromaticities identical to the ones defined in ITU-R
BT.709-3, a High-Definition TV (HDTV) standard, sRGB, as
defined in IEC 61966-2-1, strives to represent the evolution
of the standard North-American TV and its convergence with
the PC world. At the same time, its chromaticities are not very
far from the ones of SMPTE-C (and SMPTE-240M), the
present North-American TV
standard, maintaining
compatibility with the large quantity of recorded media.
sRGB’s white is defined with Illuminant D65.
Advertised as a general-purpose space for consumer use,
sRGB is proposed for applications where embedding a color
space profile, such as an ICC profile, may not be convenient
for file size or compatibility purposes. By having all elements
in a system sRGB compliant, no time is lost in conversions.
The World Wide Web is obviously a target of choice for this
space but it should not be discounted for other “scanner-to-
printer” applications. An extended gamut color encoding
standard has been defined for sRGB7; it supports multiple
levels of precision while being compatible with the base
standard.