CIE94, History, Use, and Performance
The CIE94 color-difference formula  was developed by the CIE Technical Committee 1–29 “Industrial Color Difference Evaluation” chaired by Dr. D.H. Alman (USA). CIE94 is a CIELAB-based color-difference formula where CIELAB lightness, chroma, and hue differences are appropriately weighted by very simple “weighting functions” correcting CIELAB’s lack of visual uniformity, as well as by “parametric factors” accounting for the influence of illuminating/viewing conditions in color-difference evaluation. CIE94 may be considered a simplified version of the CMC color-difference formula , as well as a predecessor of currently CIE-recommended color-difference formula, the joint ISO/CIE standard CIEDE2000 [3, 4].
In 1976, CIE recommended the CIELUV and CIELAB color spaces with their corresponding color-difference formulas, defined as Euclidean distances in such spaces. These two color spaces are only approximately uniform, and it was indicated that the use of different weighting for lightness, chroma, and hue differences may be necessary in different practical applications . CIELAB has been widely accepted in many industrial applications , and in the past recent years, several CIELAB-based color-difference equations have been proposed to improve its performance. Among these formulas, we can mention, in chronological order, JPC79 , CMC , BFD , and CIE94. Nowadays, it is generally agreed that most of these formulas significantly improved CIELAB .
Illumination: D65 source
Illuminance: 1000 lx
Observer: Normal color vision
Background field: Uniform, neutral gray with L * = 50
Viewing mode: Object
Sample size: Greater than 4 degrees
Sample separation: Direct edge contact
Sample color-difference magnitude: Lower than 5.0 CIELAB units
Sample structure: Homogeneous (without texture)
In the textile industry, it is common practice to set the lightness parametric factor k L to 2. Although the experimental conditions leading to this parametric correction to lightness difference are not yet well understood, it introduces important improvements in the performance of CIE94, which in this case must be designated as ΔE* 94 (2:1:1) or CIE94 (2:1:1).
The development of CIE94 [12, 13] began with a selection of experimental visual datasets meeting the next main conditions: statistical significance (i.e., to represent a population average with its corresponding uncertainty), well-documented experimental conditions, and use of object color specimens. CIE TC 1–29 decided to use only three experimental datasets: Witt [14, 15], Luo and Rigg , and RIT-DuPont [17, 18]. The main goal was to use these datasets to find the best weighting functions S L , S C , S H correcting CIELAB. The analyses also considered the characteristics of the previous CMC color-difference formula, which was an ISO standard in textiles  and was successfully employed in different industries.
As indicated by Eqs. 7 and 8, it was found that simple linear chroma functions described the main trends in the experimental datasets analyzed. Both the lightness dependence of lightness tolerances and the hue angle dependence of hue tolerances proposed by the CMC color-difference formula were found not robust trends in the experimental datasets analyzed , and therefore they were disregarded in CIE94. It is possible that these two corrections proposed by CMC were influenced by some specific parametric factors. Anyway, it can be said that CIE94 was a conservative approach incorporating only the main corrections to CIELAB. Currently, the CIEDE2000 color-difference formula has incorporated additional corrections to CIELAB. For example, CIEDE2000 proposes a V-shaped function for the S L function, which is different to both the lightness function proposed by CMC and the simple S L = 1 adopted by CIE94.
CIE94 was the first CIE-recommended color-difference formula incorporating the influence of illuminating/viewing experimental conditions in color-difference evaluations through the use of the so-called parametric factors k L , k C , k H (see Eq. 1).
The work carried out by CIE TC 1–29 finished with the proposal of CIE94 and some guidelines for coordinated future work on industrial color-difference evaluation . These guidelines updated those earlier given by Robertson , proposing a new set of 17 color centers to be studied in future research, considering the effects of changes from the “reference conditions,” and suggesting the development of a database of color-difference visual responses.
From the combined dataset employed at CIEDE2000 development , the performance of different color-difference formulas has been tested using the STRESS index . Low STRESS values (always in the range 0–100) indicate better color-difference formula performance. STRESS values for CIELAB, CIE94, and CIEDE2000 are 43.9, 32.1, and 27.5, respectively . As we can see, the improvement achieved by CIE94 upon CIELAB was considerably higher than the one achieved by CIEDE2000 with respect to CIE94. From STRESS values, it can be also concluded that in CIE94 the S C function (which was also adopted in CIEDE2000) is a much more important correction to CIELAB than the S H function . Besides CIE94 being proposed for object colors, it has been reported that this formula also performed satisfactorily for self-luminous color datasets .
Curiously, Eqs. 7 and 8 involve the geometrical mean of the CIELAB chroma of the two samples in the color pair, in place of the more simple arithmetical mean proposed in other color-difference formulas, for example, CIEDE2000. Strictly speaking, this implies that the locus of constant CIE94 differences with respect to a given color center is not an ellipsoid/ellipse, although deviations from ellipsoidal/elliptical contours can be considered negligible in most practical situations. CIE94 color differences computed using the geometrical and arithmetical means of the CIELAB chroma of the samples in the color pair are slightly different, in particular for colors with very low chroma .
In comparison with other recent formulas like JPC79, CMC, BFD, or CIEDE2000, it can be said that CIE94 was relevant because it was a very simple and versatile color-difference formula accounting for the main robust trends found in reliable color-difference visual datasets. CIE94 just proposes to use simple corrections to CIELAB provided by the weighting functions S L , S C , S H plus consideration of the influence of the illuminating/viewing conditions using the parametric factors k L , k C , k H . After the CIE94 proposal, CIE TC 1–47 continued further work leading to the last CIE-recommended color-difference formula, CIEDE2000, which significantly improved CIE94 for the experimental datasets used in its development .
- 1.CIE Publication 116: Industrial Colour-Difference Evaluation. CIE Central Bureau, Vienna (1995)Google Scholar
- 3.CIE Publication 142: Improvement to Industrial Colour-Difference Evaluation. CIE Central Bureau, Vienna (2001)Google Scholar
- 4.CIE S 014-6/E:2013: Colorimetry – Part 6: CIEDE2000 Colour-Difference Formula. CIE Central Bureau, Vienna (2013)Google Scholar
- 5.CIE Publication 15.2: Colorimetry, 2nd edn, Note 9, p. 33. CIE Central Bureau, Vienna (1986)Google Scholar
- 10.Brill, M.H.: Suggested modification of CMC formula for acceptability. Color Res. Appl. 17, 402–404 (1992)Google Scholar
- 12.Berns, R. S.: The mathematical development of CIE TC 1–29 proposed color difference equation: CIELCH. In: Nemcsics, A., Schanda, J. (eds.) AIC (Association Interrnationale de la Couleur). 1993. AIC Color 93, Proceedings of the 7th Congress, Budapest, 13–18 June 1993, 3 vols. Hungarian National Colour Committee, Budapest. (1993)Google Scholar
- 13.Berns, R.S.: Billmeyer and Saltzman’s Principles of Color Technology, pp. 120–121. Wiley, New York (2000)Google Scholar
- 19.ISO 105-J03: Textiles: Test for Colour Fastness – Part 3: Calculation of Colour Differences. ISO, Geneva (2009)Google Scholar
- 21.Robertson, A.R.: CIE guidelines for coordinated research on color-difference evaluation. Color. Res. Appl. 3, 149–151 (1978)Google Scholar
- 26.Melgosa, M., Huertas, R., Hita, E., Benítez, J. M.: Diferencias de color CIE94 con/sin un estímulo de referencia. In: Proceedings VI Congreso Nacional del Color (ISBN 84-699-9187-6), Area de Nutrición y Bromatología, pp. 117–118. Universidad de Sevilla (2002)Google Scholar