Paper for IS&T NIP16 Conference, Vancouver, Canada, Oct.

16-20, 2000 (Final manuscript)

CIE Fundamentals for Color Measurements

Yoshi Ohno
Optical Technology Division
National Institute of Standards and Technology
Gaithersburg, MD 20899 USA

Abstract briefly overviewed, and then practical aspects of color

The paper first overviews the CIE system of measurements and instruments – spectroreflectometers,
colorimetry, covering CIE 1931 color matching functions, spectroradiometers, and colorimeters – are discussed, with a
XYZ tristimulus values, the x, y diagram, the 1976 u', v' focus on the calibration methods and standards. Uncertainty
diagram, and the evolvement of CIELUV and CIELAB components and correction for errors are also discussed. For
color spaces and color difference formulae. The paper further details in colorimetry and color science, refer to
reviews the measurement of object colors introducing CIE official CIE publications 1-3 and many other appropriate
standard illuminants and the CIE terminology for color and references 4.
reflectance measurements, then the measurement of light-
source colors (including displays) with calculation of
correlated color temperature and color rendering indices.
CIE System of Colorimetry
The paper also discusses practical aspects of color History and Basis
measurements for imaging applications using By the early 19th century, it became known that there
spectroreflectometers, spectroradiometers, and tristimulus were three types of cones in the eyes to sense colors. It was
colorimeters. Overview is given for calibration and also known that two light stimuli having different spectra
verification of instruments’ accuracy, spectral irradiance and could produce the same color (metamerism). It was inferred
reflectance standards (available from national laboratories), that each cone had spectral sensitivities corresponding to R,
and uncertainty components. G, B (Trichromatic Theory, Young, 1800’s) or sensitivities
corresponding to opponent colors, W/Bk, R/G, and Y/B
Introduction (Opponent Theory, Hering, late 1800’s). The spectral
The term color is used with different meanings in sensitivities of the cones were yet to be known at that time,
different technologies. To lamp engineers, color refers to a but a color could be matched by combination of three
property of light sources. To graphics art engineers, color is primaries, which could be used to specify color (Maxwell,
a property of an object’s surface (under a given 1860’s).
illumination). In each case, color must be physically Around 1930, Wright and Guild made independent
measured in order to record it and reproduce the same color. visual experiments to derive color matching functions using
The perception of color is a psychophysical phenomenon, three R/G/B primaries, the results of which became the basis
and the measurement of color must be defined in such a way of the CIE colorimetry system. Observers viewed a 2°
that the results correlate accurately with what the visual circular split field and their task was to adjust the three
sensation of color is to a normal human observer. primaries so that their mixture visually matched the visible
Colorimetry is the science and technology used to quantify spectrum presented sequentially. Fig. 1 shows the results of
and describe physically the human color perception. The this experiment using a set of primaries at 435.8 nm, 546.1
basis for colorimetry was established by CIE (Commission nm, and 700 nm. This is the plot of the relative intensities of
Internationale de l clairage) in 1931 based on visual R,G,B primaries (white-balanced to equal energy white) that
experiments. Even though limitations are well recognized, matched monochromatic stimulus at each wavelength. The
the CIE system of colorimetry remains the only minus value means that one of the primary colors had to be
internationally agreed metric for color measurement. All added to the monochromatic stimulus to make the match. In
the official color-related international standards and 1931, the CIE adopted these results as the standardized
specifications use the CIE System. The CIE system works RGB color matching functions denoted as r (λ ), g (λ ), b (λ ) .
well in most cases, but one should know the assumptions Then still in 1931, the CIE transformed the RGB color
and limitations in visual conditions where the CIE system is matching functions to a new set of primaries, XYZ, to
defined. In this paper, the CIE system of colorimetry is eliminate negative values and with the g(λ ) function set to

-1-
 Tristimulus Value
By using the color matching functions, light stimuli
r (λ ) having any spectral power distribution can be specified for
2.0
g (λ ) color by three values:
b (λ )
∫
X = k Φ (λ ) x (λ )dλ
Value

1.0 λ

Y = k ∫ Φ (λ ) y (λ )dλ (1)
λ
0.0
Z = k ∫ Φ (λ )z (λ )dλ ,
λ
where Φ (λ ) is the spectral distribution of light stimulus and
-1.0
350 400 450 500 550 600 650 700 750
k is a normalizing constant. These integrated values are
called Tristimulus values. For light sources and displays,
Wavelength (nm)
Φ (λ ) is given in quantities such as spectral irradiance and
spectral radiance. If Φ (λ ) is given in an absolute unit and
Figure 1 – CIE 1931 RGB color matching functions k=683 lm/W is chosen, Y yields an absolute photometric
using primaries at 435.8 nm, 546.1 nm, and 700 nm. quantity such as illuminance or luminance.
For object colors, Φ (λ ) is given by

Φ (λ ) = E(λ ) ⋅ R(λ ) , (2)

2.0
where R(λ ) is the spectral reflectance or radiance factor of
x ( λ)
y (λ )
the object, E(λ ) is the (relative) spectral irradiance of the
Value

illumination, and
z (λ )

1.0 ∫
k = 100 / E(λ ) y (λ )dλ .
λ
(3)

Actual integration can be carried out by numerical

summation of spectral data.

0.0 Chromaticity Diagrams

350 400 450 500 550 600 650 700 750 By projecting the tristimulus values on to the unit plane
(X+Y+Z=1), color can be expressed in a two dimensional
Wavelength (nm)
plane. Such a unit plane is known as the chromaticity
diagram. The color can be specified by the chromaticity
Figure 2 – CIE 1931 XYZ color matching functions. coordinates (x, y) defined by
x= X ; y= Y . (4)
be equal to the 1924 CIE spectral luminous efficiency X +Y + Z X +Y + Z
function, V(λ). This is simply a linear transformation from
the RGB color matching functions, and the resulting The diagram using the chromaticity coordinates (x, y) is
functions, shown in Fig. 2, are called the CIE 1931 XYZ referred to as the CIE 1931 chromaticity diagram, or the
color matching functions denoted as x (λ ), y(λ ), z (λ ) . There CIE (x, y) chromaticity diagram.
are two important assumptions in these color matching The (x, y) chromaticity diagram is very non-uniform in
functions: First, rod intrusion is excluded, thus it applies terms of color difference. The minimum perceivable color
only to narrow field-of-view (2°). Second, additivity of light differences in the CIE (x, y) diagram, known as MacAdam
stimuli (Grassmann’s Law) is assumed. The ideal observer ellipses, are shown in Fig. 3(a). To improve this, in 1960,
whose color-matching properties correspond to the color CIE defined an improved diagram – CIE 1960 (u, v)
matching functions x (λ ), y(λ ), z (λ ) with the 2° field of chromaticity diagram (now deprecated), and in 1976, a
view and satisfying the Grassmann’s Law is called the CIE further improved diagram – CIE 1976 uniform chromaticity
1931 standard colorimetric observer. Practically, this scale (UCS) diagram, with its chromaticity coordinate
observer can be used for field-of-view of up to 4°. In 1964, (u’, v’) given by
the CIE defined a second set of standard color matching
4X 9Y
functions for a 10° field-of-view, denoted as u′ = ; v′ = . (5)
x 10(λ ), y 10(λ ), z 10(λ ) , to supplement those of the 1931 X + 15Y + 3Z X + 15Y + 3Z
standard observer. This is called the CIE 1964 The 1976 (u’, v’) chromaticity diagram is significantly more
supplementary standard colorimetric observer, and can be uniform than the (x, y) diagram, yet it is still far from perfect
used for a field of view greater than 4°. as shown in Fig. 3 (b).

-2-
 0.9
L*=100
yellowish (white)
b*
0.8

0.7
greenish a*
0.6

0.5
-a* reddish
y
0.4
-b*
0.3
blueish
0.2 L*=0
0.1
(black)

0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Figure 4 – CIELAB color space.
x
(a) CIE 1931 (x, y) chromaticity diagram

0.7
in the CIELAB space is calculated as the Euclidean distance
between the points in this three-dimensional space, and is
0.6 given by,
0.5 ∆E *ab = [( ∆L* )2 + ( ∆a* ) 2 + ( ∆b * )2 ]1 / 2 . (6)
* * *
0.4 This equation is called the CIE 1976 (L a b ) color
v’ difference formula. The chroma C *ab and the hue angle hab
0.3
are also calculated from (L*a*b*) by
0.2
C*ab = (a*2 + b*2)1/2 (7)
0.1
hab=tan-1 (b*/a*). (8)
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
The CIELUV space is defined in a similar manner, and
u’ the coordinate (L*,u*,v*) is calculated from the Y and (u’,v’)
(b) CIE 1976 (u’,v’) chromaticity diagram of the given light stimulus and the white point. Refer to
references 1 and 4 for the details.
While the color difference ∆E *ab is widely used, its
Figure 3 – MacAdam Ellipses in CIE 1931(x, y)
chroma scale is known to be fairly nonlinear. For more
diagram and the CIE 1976 (u’v’) diagram. The
accurate color difference evaluations, CIE recommended an
ellipses are plotted 10 times their actual size.
improved industrial color difference formula in 1994 -
CIE94 Formula 5. The color difference ∆E * 94 is calculated
from ∆L* , ∆C *ab , and ∆H *ab of the CIELAB formula.
Uniform Color Spaces and color difference formulae Another improved formula, the CMC Colour Difference
Three attributes of color, are hue, chroma (saturation), Formula, is mainly used in textile industry 6. Further
and lightness, and are expressed in a three dimensional improved color difference formulae are being investigated
space. In the chromaticity diagrams as mentioned above, by CIE (TC1-55).
lightness is missing, and the hue and chroma are laid out
very nonlinearly. To allow accurate specification of object Correlated Color Temperature
colors and color differences, CIE recommended three- The color of light sources are measured and expressed
dimensional uniform color spaces – CIELAB and CIELUV by the resultant chromaticity coordinates (x,y) or (u’,v’).
in 1976. Since the equations are long, they are omitted here. However, it is difficult to relate these values immediately to
These are called the CIE 1976 (L*a*b*) color space or
particular colors. For such practical purposes, the color of
CIELAB color space, and the other, CIE 1976 (L*u*v*) color
“white light” can be expressed by correlated color
space or CIELUV color space, and have similar structures
temperature (CCT) in the unit Kelvin [K]. The CCT is
as the Munsell color solid. In imaging applications,
defined as the temperature of the Planckian radiator whose
CIELAB space is commonly used. In CIELAB space, L *
perceived color most closely resembles that of a given
shows the lightness, and (a *, b*) the color as shown in
stimulus at the same brightness and under specified viewing
Fig. 4, The coordinate (L *,a*,b*) is calculated from the
conditions 7. For example, 2800 K is immediately associated
(X,Y,Z) of the given light stimulus and (Xn,Yn,Zn) of the
with the warm color of incandescent lamps, and 9000 K the
white point. Therefore, the CIELAB space has a function of
bluish white from a CRT. According to this definition, CCT
correcting for chromatic adaptation to the white point, and is
can be calculated using one of the chromaticity diagrams.
intended for object color and displays. The color difference

-3-
Due to the long tradition, CIE still recommends to calculate
CCT using the 1960 (u,v) chromaticity diagram (now
Measurement of Object Color
deprecated) 1. From (u’,v’) coordinates, (u,v) can be
obtained by u=u’, v=2v’/3. On the (u,v) diagram, find the Terminology of reflectance measurement
point on the Planckian locus that is at the shortest distance Object color, in most cases, is determined by spectral
from the given chromaticity point. CCT is the temperature reflectance measurements. The terminology for reflectance
of the Planck’s radiation at that point. A practical way of measurements is often confused and misused by the imaging
computing CCT is available 8. community. Some important terms are reviewed here
according to Ref. 7.
Color Rendering Index Reflectance is the ratio of the reflected radiant or
luminous flux to the incident flux in the given conditions of
For lamps in lighting applications, it is important to
spectral composition, polarization, and geometrical
evaluate how well a given illumination can render colors of
distribution. The geometrical conditions are very important
objects in the illuminated scene. The CIE defined the color
for correctly describing and measuring reflectance, and can
rendering index (CRI) for the first time in 1965. Going
lead to confusion regarding reflectance measurements.
through minor revisions, the CIE recommendation9 has been
Perfect Reflecting Diffuser is an ideal isotropic diffuser
in wide use mainly by lighting industry. The procedure of
with a reflectance equal to 1.
calculation is first to calculate the color differences ∆Ei (on
Reflectance Factor is the ratio of the radiant or
the 1964 W*U*V* uniform color space – now obsolete) of
luminous flux reflected in the direction delimited by the
selected 14 Munsell samples between the conditions when
given cone to that reflected in the same direction by a
illuminated by a reference illuminant and when illuminated
perfect reflecting diffuser identically irradiated or
by the given illumination. The process incorporates the von
illuminated.
Kries chromatic adaptation transformation. Then the Special
Radiance Factor is the ratio of the radiance of a surface
Color Rendering Index R i for each color sample is
element in a given direction to that of a perfect reflecting
calculated by
diffuser identically irradiated.
Ri = 100 – 4.6 ∆Ei . (9) Radiance Coefficient is the ratio of the radiance of the
This gives an indication of color rendering for each surface element in the given direction to the irradiance on
particular color. The General Color Rendering Index, R a, is the medium.
given as the average of the first eight color samples There are several important implications that follow
(medium saturation). With the maximum value being 100, from the above definitions. “Factor” in these terms means
Ra gives a scale that matches well with the visual impression with respect to a perfect reflecting diffuser, and therefore
of color rendering. For example, lamps having R a values can be greater than one. Reflectance, on the other hand, can
greater than 80 may be considered suitable for interior never be greater than one, and is often used descriptively to
lighting, and R a greater than 90 for visual inspection represent all of these reflectance-related quantities.
purposes. Reflectance factor is defined in terms of a cone, while
radiance factor is defined only in terms of a direction.
Standard Illuminants Therefore, if the solid angle of the cone approaches zero, the
reflectance factor approaches the radiance factor for the
The colors of objects change depending on the same conditions of irradiation. If the solid angle of the cone
spectrum of illumination. Thus, there is a need to specify approaches 2π sr, the reflectance factor approaches the
the illumination for any object color specification. For this reflectance for the same conditions of irradiation. Finally,
purpose, colorimetric illuminants are standardized by CIE radiance coefficient is similar to the bi-directional
and ISO 1, 2. CIE Standard Illuminant A (representative of reflectance distribution function (BRDF) except that the
tungsten-filament lighting with a color temperature of latter is defined for directional incident flux.
2856 K) and CIE Standard Illuminant D65 (representative
of average daylight with a CCT of 6500 K) are the two
primary standard illuminants 2. It is recommended that either Illuminating and viewing conditions
of these illuminants be used in all applications. However, Geometry is one of the most important conditions to
other phases of daylight illuminant are already widely used specify in reflectance colorimetry. For the colorimetry of
in specific application areas, and CIE also defines D50, D55 objects, CIE recommends the use of one of four standard
and D75 1. Equations are available to obtain the data table geometries – 45º/normal (45/0), normal/45º (0/45),
for Illuminant A and any phase of D illuminant. Even diffuse/normal (d/0), and normal/diffuse (0/d). The details
though no longer recommended for use, Illuminant B was on this subject are covered by the paper by Danny Rich 10.
intended to represent direct sun light with a CCT of
~4900 K, and Illuminant C to represent average daylight Reflectance standards
with a CCT of ~6800 K and to be realized by a tungsten Most spectroreflectometers are calibrated using white
source combined with a prescribed liquid filter. reflectance standards for one of the geometries listed above.
Spectral radiance factor standards are needed for the 45/0,

-4-
0/45, and d/0 geometries, while diffuse spectral reflectance upon the characteristics of the test sample. For example,
standards are needed for the 0/d geometry. Highly diffuse saturated color samples tend to have larger errors.
white materials such as pressed or sintered In order to verify stated measurement uncertainties for
polytetrafluoroethylene (PTFE) are used for such standards. spectrophotometers, calibrated color standards are used.
Because absolute measurements of radiance or reflectance Ceramic tiles of various colors manufactured by the British
factors are very difficult, calibrated standards are provided Ceramic Research Association (BCRA) are available from
by national metrology laboratories 11, 12, and industrial the National Physical Laboratory (NPL), UK and will be
measurements are normally made with traceability to these available from the National Institute of Standards and
standards. Since a perfect reflecting diffuser does not exist, Technology (NIST), USA.
the radiance factor is calibrated by absolute measurements For measurements of small color differences,
of the radiance coefficient. The radiance factor is then tristimulus colorimeters are used because of their benefits of
obtained from the radiance coefficient by multiplying by the high speed and low cost. The uncertainty of tristimulus
constant π. colorimeters is limited, however, due to the mismatch of the
The reflectance characteristics of even the most diffuse illumination to the CIE illuminants and of the spectral
materials are sensitive to the illumination and viewing response of the detectors to the CIE color matching
angles. An example for the measured spectral radiance functions. Thus, they are not suitable for absolute color
factor of a pressed PTFE sample is shown in Fig. 5. measurements over a wide range of colors 14.
A number of recommendations on spectral reflectance
and color measurements are available from the American
1.1 0° incidence Society for Testing and Materials (ASTM) 15.
1
0.9 Measurement of Light Source Color
Radiance Factor

0.8
-90 -60 -30 0 30 60 90
The measurement of light source color is represented by
measurement of lamps, LEDs, and displays. Both spectro-
1.1
radiometers and tristimulus colorimeters are widely used.
45° Incidence
1
Measurement instruments for light source color
0.9
Spectroradiometers are normally designed to measure
0.8
either spectral irradiance (unit: W m-2 nm- 1) or spectral
-90 -60 -30 0 30 60 90
radiance (unit: W sr-1 m- 2 nm-1). The former is equipped
Viewing angle (degree) with a diffuser or a small integrating sphere as input optics,
and the latter equipped with imaging optics. For example,
Figure 5 – Radiance factor of pressed PTFE as a
lamps are normally measured for spectral irradiance and
function of viewing angle at a wavelength of 555 nm. displays are measured for spectral radiance to obtain colors.
There are two types of spectroradiometers; mechanical
scanning type and diode-array type. Generally, the former is
Measurement instruments for object color more accurate but slow, and the latter is fast but less
Spectroreflectometers are commonly used for object accurate. Spectroradiometers are calibrated against spectral
color measurements. These instruments measure the irradiance or radiance standards traceable to national
spectral reflectance of a test sample under a given standards 16. Thus, their measurement uncertainty first
geometrical condition, and most are calibrated by a depends on that of the reference standard. Then, like
reference standard traceable to a national metrology spectroreflectometers, there are many other uncertainty
laboratory. Thus, their measurement uncertainty first components including wavelength error, detector
depends upon the uncertainty of the reference standard. nonlinearity, stray light of monochromator, bandwidth,
Uncertainties also arise from the characteristics of the measurement noise, etc. The errors vary depending on the
spectroreflectometer. Effects contributing to the uncertainty spectrum of the source measured. Even if the instrument’s
include wavelength error, detector nonlinearity, stray light, specification shows a low uncertainty for a tungsten source
bandwidth, the geometrical conditions for both illumination (normally a calibration source), the instrument’s uncertainty
and viewing, and measurement noise. The effect of for other colors can be much larger. For example, typical
bandwidth can be serious for bandwidths greater than diode-array spectroradiometers exhibit errors of up to 0.005
10 nm. For example, a 20 nm bandwidth can cause errors of in x, y for various display colors while they are specified for
as much as two to three CIELAB units for saturated colors. an uncertainty of ~0.001 in x, y for CIE Illuminant A. For
An effective correction method is available 1 3 . applications where highest accuracy is required, it is
Recommendations for standard methods to characterize necessary to calibrate the instruments for various actual
spectroreflectometers are being developed by the CIE TC2- colors to be measured. For color measuring instruments for
28 committee. Finally, the uncertainty can also depend displays, such a calibration facility and services are
available 17.

-5-
 Tristimulus colorimeters are also widely used for 7. CIE Publ. No. 17.4/IEC Pub. 50(845) International
colorimetry of lamps and displays. While they have benefits Lighting Vocabulary (1989).
of low cost and high speed, errors due to spectral mismatch 8. R. Robertson, Computation of correlated color
are inevitable and their uncertainty for measurement tend to temperature and distribution temperature, J. Opt. Soc.
be higher than spectroradiometers. To improve accuracy in Am. 58, 1528-1535 (1968).
display measurements, effective correction techniques are 9. CIE Publ. No. 13.3, Method of Measuring and
available 18, 19. Specifying Colour Rendering Properties of Light
Sources (1995).
10. D. Rich, Advances in the Geometric Specifications of
Uncertainty of measurements Instruments for the Measurement of Color and Color-
Uncertainty of measurement is an estimate of the range Differences, Proc., IS&T NIP16 (2000).
of values within which the true value lies. When making 11. For example, R. Barnes, E. Early, and A. C. Parr, NIST
colorimetric and photometric measurements, it is important Special Publication 250-48 Spectral Reflectance (1998).
to know the uncertainty of the results. The uncertainty in a 12. CIE Publ. No.44, Absolute Methods for Reflection
measurement depends not only on that of the measurement Measurement (1979, reprint in 1990).
instruments but also on the measurement conditions. The 13. ASTM E308-95, Standard Practice for Computing the
measurement uncertainty must be stated for official Colors of Objects by Using the CIE System (1995).
exchange of measurement results, and it must follow the 14. ASTM E1347-90, Standard Test Method for Color and
international recommendations 20. The term accuracy is no Color-Diffrence Measurement by Tristimulus
longer recommended to specify the values of uncertainty. Colorimetry (1990).
For industrial measurements, it is now recommended to use 15. ASTM Standards on Color and Appearance
an expanded uncertainty with a coverage factor k=2. See Measurement, Sixth Edition, ASTM, West
Ref. 20 for the details. Conshohoken, PA (2000).
16. For example, J. Walker, R. Saunders, J. Jackson, and D.
McSparon, NIST Special Publication 250-20 Spectral
Conclusion Irradiance Calibrations (1987).
17. S. Brown and Y. Ohno, NIST Calibration Facility for
An overview has been given for the fundamentals of the Display Colorimeters, Proc., IS&T/SPIE 11th Intn.
CIE colorimetry system and practical issues in Symposium, Electronic Imaging '99, San Jose (1999).
measurements of object color and light-source color. When 18. ASTM E 1455-96, Standard Practice for Obtaining
making measurements, one should be aware of the Colorimetric Data from a Visual Display Unit Using
uncertainty of the measurement instruments and uncertainty Tristimulus Colorimeters (1996).
components arising from the measurement conditions. Refer 19. Y. Ohno and S. Brown, Four-Color Matrix Method for
to the given references for the details in the subjects covered Correction of Tristimulus Colorimeters - Part 2, Proc.,
in this paper. IS&T Sixth Color Imaging Conference (1998).
20. ISO, Guide to the Expression of Uncertainty in
Acknowledgement Measurement, a joint publication by BIPM, IEC, ISO,
IFCC, IUPAC, IUPAP, OIML, First edition (1993).
The author thanks Danny Rich of Sun Chemical,
Edward Early and Maria Nadal of NIST for their providing
useful information and comments.
Biography
Yoshi Ohno received a Ph.D. in engineering from Kyoto
University, Kyoto, Japan in 1993. He started his career in
References photometry and colorimetry at Matsushita Electric Ind. Co.
1. CIE Publ. No.15.2, Colorimetry, Second Edition (1986). in Japan in early 1980s, and immigrated to the U.S. in 1992
(No. 15.3 to be published) to be employed by NIST. He is currently the project leader
2. ISO/CIE10526-1991, CIE standard colorimetric for Photometry at Optical Technology Division, NIST, and
illuminants (1991). recently led a project for colorimetry of displays. He serves
3. ISO/CIE10527-1991, CIE standard colorimetric as the Secretary of CIE Division 2, and chairs two CIE
observers (1991). technical committees in photometry. He is a member of CIE
4. For example, G. Wyszecki, and W. S. Stiles, Color TC1-48 (revision of CIE 15.2) and IEC TC100/TA2.
Science: Concepts and Methods, Quantitative Data and Email: [email protected]
Formulae, John Wiley and Sons, Inc., New York (1982).
5. CIE Publ. No. 116, Industrial colour difference
evaluation (1995).
6. R. McDonald, Acceptability and Perceptibility Decisions
Using the CMC Colour Difference Formula, Textile
Chemist and Colorists 20-6, pp. 31-37(1988) and Errata,
Vol. 20, No. 8, p.10 (1988).

-6-