Colour Difference Formulae: Past, Present and Future

M. Ronnier Luo
University of Leeds, Leeds, UK

Introduction
The Commission Internationale de l'Eclairage (CIE) is responsible for the development of international
colour standards. One important goal is to establish a colour difference formula that would provide objective
colour decision-making to colour-using industries, i.e. a pass/fail decision based on a single number colour
difference value (ΔE) from a standard, regardless of the colour of the standard. Over 40 colour difference
formulae have been developed since the first CIE colorimetric system.1 Table 1 lists most of them. The goal
has been growing closer via the latest CIE recommendation, the CIEDE2000 formula.2 This paper reviews
the important development according to three different periods: before 1976 (the adoption of CIELAB and
CIELUV 1), between 1976 to 2001 (the recommendation of CIEDE2000) and after 2001. In the final period,
new research areas have been identified and some recent results will be introduced.

The Formulae Developed Before 1976

Over twenty formulae were derived before 1976. They can be grouped into three families, i.e. those
derived to fit MacAdam ellipses,3 to fit the Munsell data,4 to be linearly transformed from the CIE
tristimulus colour space. The Munsell system was based on steps of equal visual perception. The spacing of
the colour samples was intensively studied by the Optical Society of America and the CIE tristimulus values
of ideally spaced samples were published in 1943.4 It can be considered the earliest colour discrimination
data and demonstrated the non-uniformity of the CIE XYZ system. The earlist Munsell based colour
difference formulae is the Nickerson's index of fading5 and the most successful formula in this family is
ANLAB.6 A series of cube root formulae were later derived to simplify the ANLAB formula which involves
a cumbersome fifth-order polynomial function. This resulted in the CIELAB colour difference formula
introduced in 1976.1

The MacAdam data3 including 24 colour centres were studied using a split field visual colorimeter. This
set of data also demonstrated the poor uniformity of the CIE XYZ system. Although a number of formulae
were developed from the data, none of these formulae are widely used now because large differences have
been found between the experimental results based on visual colorimeter and surface colours.

The formulae in the family of linear transformation from XYZ have been widely used for additive colour
mixing such as that involving coloured lights and emissive phosphor displays. Some earlier formulae were
developed including the CIE U*V*W* apace.7 In 1976, it was refined to become CIELUV.1

CIELAB and CIELUV have been widely used, mainly because it is relatively easy to relate colours as
seen with positions on the diagram. The Δ E values are calculations of the distance between the standard and
sample in these spaces. They are used for industries concerned with subtractive mixture (surface colorant)
and additive mixture of coloured light (TV), respectively.

The Formulae Developed Between 1976 and 2001

As mentioned in the last section, the formulae developed before 1976 were mainly derived to fit the
Munsell and MacAdam data. The viewing conditions applied in these experiments are very different from
those used in industries. With this in mind, many sets of experimental results were published. Most of them
were conducted using large surface samples viewed under typical industrial viewing conditions. The
medium to small colour-difference data sets show that CIELAB and CIELUV formulae do not accurately
quantify small to medium size colour differences. Of these, the important data sets in terms of larger number
of observers and sample pairs, smaller observer variations, are those accumulated by Luo and Rigg,8 RIT-
Dupont,9 Kim and Nobbs,10 Witt.11 These data sets were used to develop more advanced formulae based on
modifications to the CIELAB formula: CMC(l:c),12 BFD(l:c),13 CIE94.14 (In general, one or two of these
data sets were used to develop each formula.) All these formulae showed a much better improvement than

1
CIELAB in predicting the available data sets. However, detailed comparisons of these formulae reveal there
are large discrepancies between their structures.

With this in mind, a CIE Technical Committee (TC) 1-47 on Hue and Lightness Dependent Correction to
Industrial Colour Difference Evaluation was formed in 1998. After close collaboration between the TC
members, a new formula, named CIEDE2000, was recommended by CIE in 2001. It includes five
corrections to CIELAB: a lightness weighting function, a chroma weighting function, a hue weighting
function, an interactive term between chroma and hue differences for improving the performance for blue
colours, and a factor for re-scaling the CIELAB a* scale for improving the performance for grey colours.
The results showed that there is a considerable improvement from the more advanced formulae such as
CIE94 or CMC to CIEDE2000 for all individual data sets.

Development Since 2001

As mentioned earlier, a very good colour difference formula, CIEDE2000 was developed. This great
breakthrough is mainly due to the accumulation of many comprehensive and reliable data sets. However,
new directions in colour difference research have been identified and described below.

• Almost all of the recent efforts have been spent on the modifications of CIELAB. CIE TC1-55 was formed
to recommend a new perceptually uniform colour space from colour vision theories. Uniform colour
spaces 15 based upon a colour appearance model such as CIECAM02 16 could be an ideal solution.

• All colour difference formulae can only be used in a set of reference viewing conditions defined by the
CIE.14 It will be valuable to accumulate new data to investigate the visual effect due to variation of
viewing parameters such as illuminant, coloured background, medium, physical size, colour difference
magnitude, separation, texture, luminance level.17,18 Subsequently, a formula capable of taking into
account different viewing parameters can be derived.

• Almost all of the colour difference formulae were developed only to predict the colour difference between
a pair of large single objects/patches. More and more applications require to predict colour differences
between a pair of pictorial images. The current formula does not include necessary components to consider
spatial variations for evaluating images. There is a need to develop a formula for this purpose.19

References
1. CIE, Colorimetry, CIE Pb. 15:2004, Central Bureau of the CIE, Vienna, 2004.
2. Luo M. R., Cui G. H. and Rigg B., The development of the CIE 2000 colour difference formula, Color
Res. Appl. 26, 340-350, 2001.
3. MacAdam D.L., Visual sensitivities to color differences in daylight, J. Opt. Soc. Am., 32, 247-274,
1942.
4. Newhall S. M., Nickerson D. and Judd D. B., Final report of the O.S.A. subcommittee on spacing of the
Munsell colors, J. Opt. Soc. Am., 33, 385-418, 1943.
5. Nickerson D, The specification of color tolerance, Tex. Res., 6, 505-514,1936.
6. Adams E. Q., X-Z planes in the 1931 ICI system of colorimetry, J. Opt. Soc. Am., 32,168-173, 1942.
7. Wyszecki G., Proposal for a new colour-difference formula, J. Opt. Soc. Am., 53, 1318-1319,
1963.
8. Luo M. R. and Rigg B., Chromaticity-discrimination ellipses for surface colours, Color Res. Appl. 11,
25-42, 1986.
9. Berns R. S., Alman D. H., Reniff L, Snyder G. D. and Balonon-Rosen M. R., Visual Determination of
Suprathreshold Color-Difference Tolerances Using Probit Analysis. Col Res Appl. 16, 297-316, 1991.
10. Kim H. and Nobbs J. H., New weighting functions for the weighted CIELAB colour difference formula,
Proc. Colour 97 Kyoto, Vol. 1, 446-449 (1997).
11. Witt K., Geometric relations between scales of small colour differences. Color Res. Appl. 24, 78-92
(1999).
12. Clarke F. J. J., McDonald R., and Rigg B., Modification to the JPC79 colour-difference formula. J Soc
Dyers Col., 100, 128-132 and 281-282, 1984.

2
13. Luo M. R. and Rigg B., BFD(l:c) colour difference formula, Part I- Development of the formula, J.
Soc.Dyers Col. 103, 86-94, 1987.
14. CIE (1995), Industrial Colour-Difference Evaluation, CIE Publ.116, Central Bureau of the CIE, Vienna,
Austria.
15. Luo M. R., Cui G., Li C., Uniform Colour Spaces Based on CIECAM02 Colour Appearance Model,
Col. Res. Appl., 31, 000-000, 2006.
16. CIE. A colour appearance model for colour management systems: CIECAM02, CIE Pub. 159 (2004).
17. Cui G. H., Luo M. R., Rigg B. and Li W, Colour-difference evaluation using CRT colours. Part I: Data
gathering, Col. Res. Appl., 26, 394-402, 2001.
18. S. S. Guan and M. R. Luo, Investigation of parametric effects using large colour differences, Color Res.
Appl., 24 pp356-368 (1999).
19. X. M. Zhang and B. A. Wandell (1996), A spatial extension of CIELAB for digital color image
reproduction, Proc. of the 4th IS&T/SID Color maging Conference.

Table 1 A list of most of the colour difference formulae.

Category of Munsell Data MacAdam Data Linear transfromation from Other

formulae XYZ
Before 1976
1935 Judd
1936 Index of Fading
1937 MacAdam
1939 Balinkin
1942 JHNBS
1943 Munsell Renotation
1944 ANLAB
1946 Saunderson & Milner
1951 Godlove
1955 DIN
1958 Reilley cube root Simon-Goodwin Hunter LAB
1963 CIEU*V*W*
1965 Friele
1967 FMC-I
1969 Moton cube root
1971 MLR FMC-II
1972 MCR
1974 ΔEa OSA
1976 CIELAB CIELUV
After 1976
1978 FCM
1980 JPC79 LABHNU
1984 CMC ATD
1986 SVF
1987 BFD
1991 KC-III
1995 CIE94
1997 LCD
1999 Kuehni
2001 CIEDE2000 Oleari