A Critical Review of the

Development of the CIE1931 RGB

Color-Matching Functions

Arthur D. Broadbent*
Département de génie chimique, Université de Sherbrooke, Sherbrooke, Québec, Canada J1K 2R1

Received 12 June 2003; accepted 7 October 2003

Abstract: This article describes the development of the lights by mixtures of three monochromatic primaries of
CIE1931 chromaticity coordinates and color-matching wavelengths of 435.8, 546.1, and 700.0 nm. This article
functions starting from the initial experimental data of complements that of Fairman, Brill, and Hemmendinger by
W. D. Wright and J. Guild. Sufficient information is given to providing a critical review of how the CIE1931 RGB chro-
allow the reader to reproduce and verify the results ob- maticity coordinates and color-matching functions were ini-
tained at each stage of the calculations and to analyze tially calculated from the combined experimental measure-
critically the procedures used. Unfortunately, some of the ments of Wright and of Guild. These authors actually
information required for the coordinate transformations presented their color-matching data in terms of chromaticity
was never published and the appended tables provide likely coordinates whose sum is automatically unity. The values of
versions of that missing data. © 2004 Wiley Periodicals, Inc. Col these could be combined with the photopic luminous effi-
Res Appl, 29, 267–272, 2004; Published online in Wiley InterScience cacy, V(␭), to calculate the appropriate color-matching
(www.interscience.wiley.com). DOI 10.1002/col.20020 functions.
This article is an attempt to reproduce the steps taken by
Key words: CIE system; CIE1931 RGB system; colorimetry,
the founders of the 1931 CIE RGB colorimetric system.
color coordinate transformations
This is still the basis for color measurement using a small
visual angle (2o), despite its shortcomings, particularly the
INTRODUCTION frequent lack of agreement with actual observations. The
procedure is set out in seven distinct stages. For each one,
In 1997 Fairman, Brill, and Hemmendinger1 reviewed the the corresponding chromaticity coordinates are identified by
principles involved in developing the CIE1931 standard a subscript indicating the stage number, for example, r4(␭),
observer. These data consist of spectral chromaticity coor- g4(␭), b4(␭) for stage 4. As far as possible, the appropriate
dinates, and corresponding color-matching functions, based coordinate transformation equations are given. The reader
on trichromatic color matching of spectral lights as a func- can thus actually carry out the various conversions leading
tion of their wavelengths from 380 to 780 nm. To avoid to the CIE1931 RGB system, starting from the initial ex-
negative values of the coordinates, the color of a spectral perimental results, and then compare the data with pub-
light must be reproduced by imaginary mixing of unreal lished values.2– 6 The results and their extension, leading to
primaries designated X, Y, and Z. In their description of the the 1931 CIE standard observer, did indeed have far-reach-
CIE1931 system, the above authors did not actually proceed ing implications for colorimetry and its applications, but the
from the original measurements of Wright2– 4 and of Guild,5 reader will see that the initial stages of the story are not
but rather from the CIE1931 RGB color-mixture data, without some confusing passages and errors.
which were originally calculated from the combined
Wright–Guild data and not determined experimentally. The
CIE RGB data are based on matching the colors of spectral WRIGHT’S COLOR-MATCHING DATA

Wright—Stage 1
*Correspondence to: Arthur D. Broadbent, Department of Chemistry,
Bishop’s University, Lennoxville, Québec, Canada J1M 1Z7 (e-mail:
Wright carried out trichromatic color-matching experiments
abroadbe@ubishops.ca) with 10 observers having normal color vision. A square
© 2004 Wiley Periodicals, Inc. visual field was used that subtended an angle of approxi-

Volume 29, Number 4, August 2004 267

TABLE I. Proposed original spectral chromaticity co- values of r1(␭), g1(␭), and b1(␭) are given in Table I and
ordinates, r1(␭), g1(␭), and b1(␭), based on W. D. match Wright’s published figure (Fig. 5 in reference 2).
Wright’s monochromatic primaries and corresponding
to Figure 5 in Reference 2.
Wright—Stage 2
Wavelength (nm) r1(␭) g1(␭) b1(␭)
Wright selected his units for the experimental primaries so
400 0.037 ⫺0.058 1.021 that differences in the degree of macular pigmentation
410 0.037 ⫺0.057 1.020 among the observers would result in each individual having
420 0.035 ⫺0.056 1.021
430 0.032 ⫺0.052 1.020
a unique set of chromaticity coordinates for the match of the
440 0.026 ⫺0.044 1.019 standard white light but all observers having essentially the
450 0.015 ⫺0.029 1.014 same values for the chromaticity coordinates of the spectral
460 0.000 0.000 1.000
470 ⫺0.021 0.066 0.955
colors. The standard white was produced using a standard-
480 ⫺0.062 0.203 0.859 ized NPL (National Physics Laboratory, UK) lamp, trans-
490 ⫺0.106 0.440 0.665 mitted through a solution filter and then reflected from a
500 ⫺0.135 0.754 0.381
510 ⫺0.117 0.952 0.165
magnesium oxide screen. The r1(␭), g1(␭), and b1(␭) coef-
520 ⫺0.064 1.013 0.051 ficients in Table I, based on his original system of units,
530 0.000 1.000 0.000 were then transformed into a set based on the same prima-
540 0.076 0.945 ⫺0.021
550 0.151 0.876 ⫺0.027
ries, but with their units adjusted so that equal amounts of
560 0.244 0.786 ⫺0.030 the red, blue, and green primaries matched the NPL stan-
570 0.352 0.672 ⫺0.024 dard white. According to Wright, the matching of the stan-
580 0.473 0.547 ⫺0.020
590 0.603 0.409 ⫺0.012
dard white is given by the following colorimetric equation:
600 0.717 0.292 ⫺0.009 NPL White ⫽ 0.243关R共650兲兴 ⫹ 0.410关G共530兲兴
610 0.813 0.193 ⫺0.007
620 0.885 0.121 ⫺0.005 ⫹ 0.347关B共460兲兴 ⫽ 0.333关R⬘共650兲兴 ⫹ 0.333关G⬘共530兲兴
630 0.943 0.061 ⫺0.004
640 0.979 0.023 ⫺0.003 ⫹ 0.333关B⬘共460 兲兴, (2)
650 1.000 0.000 0.000
660 1.010 ⫺0.010 0.000 where [R(650)], [G(530)], and [B(460)] represent one
670 1.015 ⫺0.015 0.000
680 1.019 ⫺0.019 0.000
trichromatic unit of his monochromatic primaries in the old
690 1.022 ⫺0.022 0.000 units, and [R⬘(650)], [G⬘(530)], and [B⬘(460)] one trichro-
700 1.026 ⫺0.026 0.000 matic unit of the same primaries in the new units. Thus, for
any monochromatic spectral light C of wavelength ␭, we
obtain the following:
mately 2° at the eye. It was divided horizontally into two C共 ␭ 兲 ⫽ r 1共 ␭ 兲关R共650兲兴 ⫹ g 1共 ␭ 兲关G共530兲兴
equal rectangles, one illuminated by the spectral light to be
⫹ b 1共 ␭ 兲关B共460兲兴 ⫽ r 2共 ␭ 兲关R⬘共650兲兴
matched and the other by a mixture of the three monochro-
matic primary lights. The units of these were such that equal ⫹ g 2共 ␭ 兲关G⬘共530兲兴 ⫹ b 2共 ␭ 兲关B⬘共460兲 兴 (3)
amounts of red (650 nm) and green (530 nm) matched a
yellow monochromatic light and equal amounts of green where according to Eq. (2)
(530 nm) and blue (460 nm) matched a blue-green mono- 0.333 r 1共 ␭ 兲 0.333g 1共 ␭ 兲
chromatic light. The results were represented in the usual r 2共 ␭ 兲 ⫽ , g 2共 ␭ 兲 ⫽ ,
0.243 d 0.410 d
manner [Eq. (1)] for matching one trichromatic unit of the
spectral light C of wavelength ␭ as follows: 0.333 b 1共 ␭ 兲
b 2共 ␭ 兲 ⫽ and
0.347 d
C共 ␭ 兲 ⫽ r 1共 ␭ 兲关R共650兲兴 ⫹ g 1共 ␭ 兲关G共530兲兴
0.333 r1 共␭兲 0.333 g1 共␭兲 0.333 b1 共␭兲
⫹ b 1共 ␭ 兲关B共460兲兴, (1) d⫽ ⫹ ⫹
0.243 0.410 0.347
where r1(␭), g1(␭), and b1(␭) are the scalar chromaticity Because of Wright’s deliberate selection of units, the chro-
coordinates, whose sum is de facto unity, and [R(650)], maticity coordinates for the NPL white of the 36 observers
[G(530)], and [B(460)] represent respectively one trichro- examined varied considerably because of variations in the
matic unit of each of the primaries (red, green, and blue) degree of macular pigmentation. Wright calculated the av-
with the indicated wavelengths in nanometers. The pub- erage values (r ⫽ 0.243, g ⫽ 0.410, b ⫽ 0.347) using an
lished data are in the form of graphs of r1(␭), g1(␭), and arbitrary graphical technique. Because of the spread (values
b1(␭) as a function of wavelength (every 10 nm from 400 to were in the ranges r ⫽ 0.15 – 0.35 and g ⫽ 0.35 – 0.55), the
700 nm), for the 10 individual observers, and of the aver- standard deviations for the averaged values were consider-
aged values. In his articles, Wright2– 4 gave no tabulated data able. In fact, Wright7 has expressed surprise at the precision
for this original set of chromaticity coordinates. I derived to which the 1931 CIE standard observer chromaticity co-
these by working back from those discussed in the next ordinates are now specified given the uncertainty of the
section for which Wright did provide a table of values.3 The original measurements.

268 COLOR research and application

Wright—Stage 3 values of the chromaticity coordinates in Stage 1, of the
actual equations used by Wright for the transformation in
The chromaticity coordinates derived in Stage 2 were
Stage 3, and of the precision of his manual matrix calcula-
then transformed again into a set based on the NPL
tion.
monochromatic primaries (wavelengths 700.0, 546.1, and
Wright calculated the relative luminances L for one
435.8 nm), with their units based on equal amounts of the
trichromatic unit of each monochromatic NPL primary and
red, blue, and green primaries matching the same NPL
found LR : LG : LB ⫽ 1.00 : 4.329 : 0.0468. By his own
standard white. The transformation is based on defining
admission, his data on the relative luminances of his original
the chromaticity coordinates of the monochromatic NPL
primaries were not overly precise, showing considerable
primaries in terms of Wright’s primaries. I did this using
scatter for the different observers. These relative luminances
computer-generated curves by graphical interpolation at
therefore differ from those of Guild for the same system
the exact wavelengths of the NPL primaries using fourth-
(see “Guild—Stage 5”).
order polynomials over 5 to 6 points around the wave-
length in question. The interpolated values obtained were
then adjusted slightly to minimize the differences be- GUILD’S COLOR-MATCHING DATA
tween the calculated and Wright’s published values3
(maximum adjustment ⫾ 0.001). This was necessary Guild—Stage 4
because Wright did not give the required transformation Guild’s colorimeter also had a square visual field sub-
equations as follows: tending an angle of 2o at the observer’s eye and split into
R共700兲 ⫽ 1.0510关R⬘共650兲兴 ⫺ 0.0150关G⬘共530兲兴 two equal rectangles. In other respects, his instrument
was quite different than the one used by Wright. Guild5
G共546兲 ⫽ 0.1940关R⬘共650兲兴 ⫹ 0.8338关G⬘共530兲兴 measured the average spectral chromaticity coordinates
r4(␭), g4(␭), and b4(␭) for seven observers using red,
⫺ 0.0282 关B⬘共460兲兴 (4)
blue, and green filtered lights as primaries, with units
B共436兲 ⫽ 0.0406关R⬘共650兲兴 ⫺ 0.0407关G⬘共530兲兴 such that equal amounts matched the standard NPL
white. The relative luminances of these filtered primaries
⫹ 0.9985 关B⬘共460兲兴 were LR : LG : LB ⫽ 1.00 : 2.858 : 0.169. These values of
LR, LG, and LB were initially measured but later slightly
Inversion of the 3 ⫻ 3 matrix of the coefficients from Eq.
adjusted so that summation over the entire wavelength
(4) and normalization (so that the sums of one unit of the
range gave the following:
three primaries are the same for both sets) give the speci-
fication of Wright’s primaries in terms of those of the NPL
as follows: ⌺ 冉 r 4共 ␭ 兲 E共 ␭ 兲V共 ␭ 兲
冊
r 4共 ␭ 兲 L R ⫹ g4 共␭兲 LG ⫹ b4 共␭兲 LB
R⬘共650兲 ⫽ 1.3978关R共700兲兴 ⫹ 0.0140 关G共546兲兴
⫹ 0.0005关B共436兲兴
⫽⌺ 冉 g4 共␭兲 E共␭兲V共␭兲
r4 共␭兲 LR ⫹ g4 共␭兲 LG ⫹ b4 共␭兲 LB 冊
G⬘共530兲 ⫽ ⫺ 0.3276关R共700兲兴 ⫹ 0.9479关G共546兲兴
⫹ 0.0326关B共436兲兴 (5)
⫽⌺ 冉 b4 共␭兲 E共␭兲V共␭兲
冊
r4 共␭兲 LR ⫹ g4 共␭兲 LG ⫹ b4 共␭兲 LB
(7)

as required for an achromatic white source (NPL standard

B⬘共460兲 ⫽ ⫺ 0.0702关R共700兲兴 ⫹ 0.0381关G共546兲兴 white) having its three chromaticity coordinates equal to
⫹ 0.9669关B共436兲兴 0.333. In Eq. (7) E(␭) is the relative emitted energy and V(␭)
the photopic luminous efficacy. The common denominator
Equation (5) then allows calculation of the spectral chro- in Eq. (7) is the relative luminance of one trichromatic unit
maticity coordinates r3(␭), g3(␭), and b3(␭) based on the of a spectral color of wavelength ␭.
NPL primaries as follows:
1.3978 r 2共 ␭ 兲 ⫺ 0.3276 g 2共 ␭ 兲 ⫺ 0.0702 b 2共 ␭ 兲 Guild—Stage 5
r 3共 ␭ 兲 ⫽
1.4123 r 2共 ␭ 兲 ⫹ 0.6529 g 2共 ␭ 兲 ⫹ 0.9348 b 2共 ␭ 兲
The coordinates r4(␭), g4(␭), and b4(␭) were then trans-
0.0140 r 2共 ␭ 兲 ⫹ 0.9479 g 2共 ␭ 兲 ⫹ 0.0381 b 2共 ␭ 兲
g 3共 ␭ 兲 ⫽ formed into a new set based on matching with the NPL
1.4123 r 2共 ␭ 兲 ⫹ 0.6529 g 2共 ␭ 兲 ⫹ 0.9348 b 2共 ␭ 兲 monochromatic primaries (700.0, 546.1, and 435.8 nm)
0.0005 r 2共 ␭ 兲 ⫹ 0.0326 g 2共 ␭ 兲 ⫹ 0.9669 b 2 共 ␭ 兲 with their units such that equal amounts matched the stan-
b 3共 ␭ 兲 ⫽
1.4123 r 2共 ␭ 兲 ⫹ 0.6529 g 2共 ␭ 兲 ⫹ 0.9348 b 2共 ␭ 兲 dard NPL white. The transformation equations for this are
(6) based on specifying the chromaticity coordinates of the
NPL monochromatic primaries in terms of the red, green,
The calculated values differed slightly from those published and blue filtered lights [RF], [GF], and [BF]. These equations
by Wright (only 5 of 93 values ⬎ 0.001). The deviations are given here as follows because of a printing error in
probably arise from ignorance of the actual experimental Guild’s article5:

Volume 29, Number 4, August 2004 269

R共700兲 ⫽ 1.0604关R F兴 ⫺ 0.0604关G F兴 white, the standard white used in the CIE specifications. For
the combined Wright–Guild data, Guild established the
G共546兲 ⫽ 0.0600关R F兴 ⫹ 0.9552关G F兴 ⫺ 0.0152关B F兴 (8) following:
B共436兲 ⫽ 0.04925关R F兴 ⫺ 0.06385关G F兴 ⫹ 1.0146关B F兴 Equal Energy White ⫽ 0.301关R共700兲兴 ⫹ 0.314关G共546兲兴
Inversion of the 3 ⫻ 3 matrix of the numerical coefficients ⫹ 0.385关B共436兲兴 ⫽ 0.333关R⬘共700兲兴 ⫹ 0.333关G⬘共546兲兴
in Eq. (8) and normalization (as under Wright—Stage 3)
⫹ 0.333 关B⬘共436 兲兴, (10)
give the specification of the red, green, and blue filtered
primaries in terms of the NPL monochromatic primaries where [R(700)], [G(546)], and [B(436)] represent one
that are then used to calculate the new set of spectral trichromatic unit of the NPL monochromatic primaries in
chromaticity coordinates [Eq. (9)] as follows: the old units and [R⬘(700)], [G⬘(546)], and [B⬘(436)] one
trichromatic unit of the same primaries in the new units.
1.1315 r 4共 ␭ 兲 ⫺ 0.0720 g 4共 ␭ 兲 ⫺ 0.0595 b 4共 ␭ 兲 Thus, for any monochromatic spectral light, we obtain the
r 5共 ␭ 兲 ⫽
1.1834 r 4共 ␭ 兲 ⫹ 0.8387 g 4共 ␭ 兲 ⫹ 0.9779b 4共 ␭ 兲 following:
0.0510 r 4共 ␭ 兲 ⫹ 0.8952 g 4共 ␭ 兲 ⫹ 0.0539 b 4共 ␭ 兲
g 5共 ␭ 兲 ⫽ C共 ␭ 兲 ⫽ r 6共 ␭ 兲关R共700兲兴 ⫹ g 6共 ␭ 兲关G共546兲兴 ⫹ b 6共 ␭ 兲
1.1834 r 4共 ␭ 兲 ⫹ 0.8387 g 4共 ␭ 兲 ⫹ 0.9779 b 4共 ␭ 兲
0.0009 r 4共 ␭ 兲 ⫹ 0.0156 g 4共 ␭ 兲 ⫹ 0.9835 b 4共 ␭ 兲 ⫻ 关B共436兲兴 ⫽ r 7共 ␭ 兲关R⬘共700兲兴 ⫹ g 7共 ␭ 兲关G⬘共546兲兴
b 5共 ␭ 兲 ⫽
1.1834 r 4共 ␭ 兲 ⫹ 0.8387 g 4共 ␭ 兲 ⫹ 0.9779 b 4共 ␭ 兲 ⫹ b 7共 ␭ 兲关B⬘共436 兲兴 (11)
(9) where from Eq. (10) we obtain the following:
In his article, Guild gives a table of values of r5(␭), g5(␭), 0.333 r 6共 ␭ 兲 0.333 g 6共 ␭ 兲
and b5(␭), from 380 to 780 nm, with values every 5 nm. r 7共 ␭ 兲 ⫽ , g 7共 ␭ 兲 ⫽ ,
0.301 d 0.314 d
Much of this data (45%) are interpolated, or extrapolated
below 400 and above 700 nm, using a graphical method that 0.333b 6共 ␭ 兲
placed all points at the appropriate position on the smooth b 7共 ␭ 兲 ⫽ and
0.385 d
spectral locus passing through the experimental points on
the chromaticity diagram. 0.333 r6 共␭兲 0.333 g6 共␭兲 0.333 b6 共␭兲
d⫽ ⫹ ⫹ .
The relative luminances of the NPL monochromatic pri- 0.301 0.314 0.385
maries, LR ⫽ 1.00, LG ⫽ 4.4102, LB ⫽ 0.0520, were When I transformed the original Guild–Wright spectral
calculated from those for the filtered primaries and satisfy chromaticity coordinates into the CIE RGB set using Eq.
the corresponding version of Eq. (7) with r ⫽ g ⫽ b ⫽ (11) for the new units, the results were close but not iden-
0.333. Similar summations of the terms in Eq. (7) using the tical with the CIE RGB data. The average deviations for the
simple spectral power distribution for the equal energy chromaticity coordinates (from 380 to 700 nm) were
white, with E(␭) ⫽ 1.00 at all wavelengths, gave its chro- 0.0039, 0.0044, and 0.0013 for r7(␭), g7(␭), and b7(␭)
maticity coordinates as r ⫽ 0.300, g ⫽ 0.312, and b ⫽ 0.388 respectively. Some values deviated by as much as 0.02 to
for the same primaries. 0.05, particularly around 500 nm. How could these discrep-
ancies be explained?
Guild—Stage 6 In a long article in 1932, Smith and Guild8 described how
the CIE1931 RGB system data were calculated. A key
Once Wright had calculated and published3 his chromaticity sentence appears on page 79 of their article: “Alterations [to
coordinates in terms of the same NPL primaries used by the averaged Wright–Guild spectral chromaticity co-ordi-
Guild in Stage 5, Guild realized that the two sets of data nates] were called for owing to a numerical error in one of
were so similar that it would be possible to specify an Wright’s figures, and to other irregularities present in the
average response of the eye for color matching. He calcu- unsmoothed mean of Guild’s and Wright’s results.” The
lated the numerical average of the individual chromaticity error in Wright’s data was not identified but is presumably
coordinates of Wright and Guild. The final table with coef- at or around 500 nm. The smoothing technique used by
ficients for all wavelengths from 380 to 780 nm, at a 5-nm Guild is presumably that described in his first article.5
interval, is given in Guild’s article.5 Values of chromaticity coordinates were interpolated at
1-nm intervals. On large-scale chromaticity plots, the best
curve was drawn through the experimental points. Then any
CONVERSION OF THE AVERAGED WRIGHT–GUILD
deviant interpolated points that were not on this curve were
DATA INTO THE CIE 1931 RGB SYSTEM—STAGE 7
laterally displaced onto the curve. Finally, the spacing of the
Guild’s table of averaged chromaticity coordinates5 (see points with respect to varying wavelength was regularized.
“Guild—Stage 6”) is based on equal amounts of the NPL This smoothing of the averaged Wright–Guild data was
monochromatic primaries matching the NPL standard presumably done after the publication of the combined data5
white. In this final stage, the units must be converted so that but prior to the 1931 CIE meeting in Cambridge. No other
equal amounts of the same primaries match the equal energy publication mentions this smoothing process and no table of

270 COLOR research and application

TABLE II. Proposed spectral chromaticity coordinates, r7(␭), g7(␭), and b7(␭), based on the smoothed Wright–
Guild data that were transformed into the CIE1931 RGB values.
Wavelength (nm) r7(␭) g7(␭) b7(␭) Wavelength (nm) r7(␭) g7(␭) b7(␭)

380 0.0214 ⫺0.0094 0.9881 545 ⫺0.0281 1.0267 0.0013

385 0.0210 ⫺0.0093 0.9883 550 0.0939 0.9092 ⫺0.0031
390 0.0206 ⫺0.0093 0.9887 555 0.2054 0.7995 ⫺0.0050
395 0.0201 ⫺0.0092 0.9892 560 0.3078 0.6978 ⫺0.0056
400 0.0194 ⫺0.0092 0.9898 565 0.4016 0.6039 ⫺0.0055
405 0.0186 ⫺0.0091 0.9905 570 0.4874 0.5176 ⫺0.0050
410 0.0176 ⫺0.0089 0.9913 575 0.5654 0.4389 ⫺0.0043
415 0.0162 ⫺0.0085 0.9923 580 0.6357 0.3678 ⫺0.0035
420 0.0142 ⫺0.0077 0.9935 585 0.6988 0.3041 ⫺0.0029
425 0.0111 ⫺0.0062 0.9951 590 0.7544 0.2480 ⫺0.0024
430 0.0069 ⫺0.0039 0.9970 595 0.8025 0.1994 ⫺0.0019
435 0.0009 ⫺0.0006 0.9996 600 0.8423 0.1592 ⫺0.0015
440 ⫺0.0066 0.0039 1.0027 605 0.8757 0.1254 ⫺0.0011
445 ⫺0.0166 0.0098 1.0069 610 0.9025 0.0986 ⫺0.0010
450 ⫺0.0304 0.0177 1.0127 615 0.9238 0.0770 ⫺0.0008
455 ⫺0.0480 0.0279 1.0201 620 0.9403 0.0603 ⫺0.0006
460 ⫺0.0704 0.0417 1.0286 625 0.9533 0.0472 ⫺0.0005
465 ⫺0.0988 0.0613 1.0375 630 0.9636 0.0367 ⫺0.0004
470 ⫺0.1399 0.0941 1.0458 635 0.9719 0.0283 ⫺0.0003
475 ⫺0.1976 0.1467 1.0509 640 0.9789 0.0214 ⫺0.0003
480 ⫺0.2794 0.2308 1.0486 645 0.9844 0.0158 ⫺0.0003
485 ⫺0.3952 0.3619 1.0333 650 0.9884 0.0118 ⫺0.0001
490 ⫺0.5446 0.5555 0.9891 655 0.9915 0.0086 ⫺0.0001
495 ⫺0.7274 0.8215 0.9060 660 0.9938 0.0064 ⫺0.0001
500 ⫺0.9158 1.1360 0.7798 665 0.9952 0.0049 ⫺0.0001
505 ⫺1.0632 1.4457 0.6175 670 0.9965 0.0036 ⫺0.0001
510 ⫺1.1189 1.6852 0.4337 675 0.9974 0.0026 0.0000
515 ⫺1.0502 1.7859 0.2644 680 0.9983 0.0017 0.0000
520 ⫺0.8818 1.7331 0.1487 685 0.9991 0.0009 0.0000
525 ⫺0.6771 1.5922 0.0849 690 0.9996 0.0004 0.0000
530 ⫺0.4805 1.4331 0.0474 695 0.9999 0.0001 0.0000
535 ⫺0.3114 1.2874 0.0240 700 1.0000 0.0000 0.0000
540 ⫺0.1623 1.1527 0.0096 780 1.0000 0.0000 0.0000

corrected data is available for the required transformation are LR ⫽ 1.00, LG ⫽ 4.4399, and LB ⫽ 0.0472. Note that
into the CIE RGB system. Eq. (12) is slightly different than 10, as are the values of the
I derived this “smoothed” set of chromaticity coordinates relative luminances.
by working backwards from the CIE RGB data using the The CIE RGB data consist of a set of spectral chroma-
reverse transformation i.e., the CIE1931 RGB chromaticity ticity coordinates [r(␭), g(␭), and b(␭)] and color-matching
coordinates, based on equal amounts of the NPL monochro- functions r៮(␭), g៮ (␭) and b៮ (␭). It can be exactly reproduced
matic primaries matching the equal energy white, were using the chromaticity coordinates in Table II and Eqs. (11)
transformed into a set based on the same primaries but with and 12. The individual sums of the CIE RGB color-match-
units such that the match for the NPL white satisfied Eq. (7). ing functions over all wavelengths have equal values. The
The set of coefficients obtained differs slightly from that color-matching functions are calculated from the chroma-
published by Guild and is given in Table II. In fact, several ticity coordinates using the photopic luminous efficacy V(␭)
iterative backward and forward transformations were re- and satisfy the following equation:
quired to give the exact set that would give the exact CIE
data. Each time a modified set of Guild–Wright coefficients r៮ 共 ␭ 兲 L R ⫹ g៮ 共␭兲 LG ⫹ b៮ 共␭兲 LB ⫽ V共␭兲 (13)
were obtained, the chromaticity coordinates for the equal at any wavelength ␭ for LR ⫽ 1.000, LG ⫽ 4.5907, and LB
energy white had to be modified by normalizing the three ⫽ 0.0601.
equations in Eq. (7). Finally, the coefficients in Table II
were obtained for which we have the following:
CONCLUSIONS
Equal energy white ⫽ 0.301关R共700兲兴 ⫹ 0.314关G共546兲兴
This article has presented in some detail all the steps re-
⫹ 0.385关B共436兲兴 ⫽ 0.333关R⬘共700兲兴 ⫹ 0.333关G⬘共546兲兴 quired to derive the CIE1931 RGB spectral chromaticity
⫹ 0.333关B⬘共436 兲兴. (12) coordinates from the original experimental color-matching
data. The procedures used may not correspond exactly to
The coefficients in Table II differ slightly from those in those actually used by Wright and Guild because of the lack
Table IV of reference 5. The relative luminances of the NPL of some required information in the original publications.
monochromatic primaries [R(700)], [G(546)], and [B(436)] The results given here were generated using a computer

Volume 29, Number 4, August 2004 271

spreadsheet, and although they cast a shadow on the preci- 3. Wright WD. A re-determination of the mixture curves of the spectrum.
sion of the 1931 specification, there is no doubt that the Trans Opt Soc London 1929 –30;31:201–211.
4. Wright WD. A re-determination of the trichromatic mixture data. Med
original data were more than adequate for color-matching
Res Council Spec Rep Ser 1930;No.139:5–38.
purposes. Today, the calculations and the “smoothing” of 5. Guild J. The colorimetric properties of the spectrum. Philos Trans Roy
the data would be performed quite differently, but this is not Soc London Ser. A 1931;230:149 –187.
a criticism because the CIE1931 standard observer, derived 6. CIE Publication No. 15.2. Colorimetry. 2nd ed. Vienna: Central Bureau
from the RGB data, has proved its value. In fact, later of the CIE; 1986. p 66 – 67.
measurements by Stiles and Burch,9,10 although slightly 7. Wright WD, The origins of the 1931 CIE system. In: Kaiser PK,
Boynton RM. Human color vision. 2nd ed. Washington DC: Optical
different in the blue-violet region of the spectrum, were not Society of America; 1996.
sufficiently different from those Wright and Guild to merit 8. Smith T, Guild J. The CIE colorimetric standards and their use. Trans
any change in the CIE specification. More than 70 years after Opt Soc London 1931–32;33:73–134.
the fact, it is not possible to verify the data and equations not 9. Stiles WS. The basic data of colour matching. In: Physical Society year
included in the original articles. Perhaps these omissions were book. London: Physical Society; 1955. p 44 – 65.
10. Stiles WS, Burch JM. Interim report to the CIE, Zurich 1955, on the
a consequence of the urgency to finalize the data in preparation
National Physics Laboratory’s investigation of colour matching. Op-
for the 1931 CIE meeting in Cambridge that has been de- tica Acta 1955;2:168 –176.
scribed by Wright.11 This article has been an attempt to fill 11. Wright WD. The historical and experimental background to the 1931
those gaps. Despite all the shortcomings of the 1931 CIE CIE system of colorimetry. In: Golden jubilee of colour in the CIE.
system, it is still used and still undergoing evaluation.12–14 Bradford, UK: Society of Dyers and Colourists; 1981. p 3–18.
12. Fairchild MD. A novel method for the determination of color matching
1. Fairman HS, Brill MH, Hemmendinger H. How the CIE1931 color- functions. Color Res Appl 1989;14:122–130.
matching functions were derived from the Wright–Guild data. Color 13. North AD, Fairchild MD. Measuring color matching functions. Part I.
Res Appl 1997;22:11–23. Color Res Appl 1993;18:155–162.
2. Wright WD. A re-determination of the trichromatic coefficients of the 14. Shaw M, Fairchild MD. Evaluating the 1931 CIE color matching
spectral colours. Trans Opt Soc London 1928 –29;30:141–164. functions. Color Res Appl 2002;27:316 –329.

272 COLOR research and application