OPTICAL REVIEW Vol. 8, No. 2 (2001) 142-147 Estilnation of Color-Difference Formulae at Color Discrimination Threshold Using CRT-Generated Stimuli Haisong XU,1,2 Hirohisa YAGUCH11 and Satoshi SHIOIR11 iDepartment of 1leformation aud Image Sciences. Chiba. Ultiversity, I -33 Yayoi-cho. Inage-ku. Chiba 263-8522 Japan, 2State Key Laboratory ofModern Optical 11cstrumentation. Zhejiang University. Hangzhou 310027 Chilea (Received December 26, 2000; Accepted February 14, 2001) The advanced color-difference formulae, CMC, CIE94, and the recently proposed CIEDE2000 formula, together with the basic CIELAB system, were estimated using the chromaticity discrimination threshold data at CIE Gray and Blue color centers. Gray is the most basic color and blue is, perceptually, in the most different region from other areas across the color space. The test stimuli, evenly distributed in the (a*, b*)-, (a*, L*)-, and (b*, L*)-plane of the CIELAB space, were generated on a CRT display, driven by a VSG system. Each direction from each color center was assessed 3 times by a panel of 9 normal color-vision observers with the psychophysical method of interleaved staircase. The experimental data were reliable and consistent with other studies according to the observer accuracy and fitted ellipse parameters. A comprehensive analysis shows the color discrimination tolerances could be well fitted by ellipses, and the CIEDE2000 and CIELAB formulae performed better than CIE94 while the CMC worst at the threshold level for the color centers studied. Key words: color difference, color-difference formula, color discrimination threshold, CRT, method of staircase, chromaticity discrimination ellipse, CIELAB color space 1. Introduction Since the CIELAB spacel) was recommended by the In- ternational Commission on 11lumination (CIE) in 1976, for promoting uniformity of color-difference practice, sig- nificant advances have been achieved in this field . Alon- gside the development in color vision research, new ad- vanced color-difference formulae based on CIELAB have been proposed, aiming at a final goal2) of developing an universal standard of color-difference evaluation for most industrial applications. Among these, CMC,3) CIE94,4) and the newest CIEDE2000,5) proposed recently by CIE TC1-47 Hue and Lightness-Dependent Correc- tion to Industrial Color-Difference Evaluation, are the representative models. A11 these color-difference formu- lae were developed to fit some experimental data sets. Therefore, their performances, in general, need to be tested with different data sets for practical application and further improvement. An ideal color-difference formula is expected to per- form well for the range from small, through moderate, to large color differences. In practice, it is very diflicult for a single formula to give out the color-difference metrics equal to perceptual scales over the whole range. At least such a color-difference formula has not yet been achieved. However, different formulae can meet the needs of different applications, with respect to a specified range of color differences, with the level of color discrimi- nation threshold being the cornmon basis for all of these color-difference models. On the other hand, the CRT is being widely used6-8) in vision research not only because of its efliciency, saving E-mail: hsxu@vision.tp.chiba-u.ac.jp of time, Iabor, and cost needed to produce object-color samples, but also its flexibility; it makes easy the selec- tion of test stimuli and makes it possible to study the parametric effects on color-difference judgments. The main purpose of the present study was to estimate the performances of the advanced color-difference formu- lae, CMC, CIE94, and CIEDE2000, together with the basic CIELAB system, at the color discrimination threshold. The chromaticity discrimination data were obtained from a psychophysical experiment carried out in CIELAB color space using CRT-generated stimuli, described below. 2. Methods 2. I Apparatus aud Stimuli A CRT monitor of Sony Multiscan G500 was used for presentation of color stimuli. The CRT was driven by a visual stimulus generator system, Cambridge Research Systems VSG 2 /4, with 15-bit resolution. Observations were made in a booth in which the environment was black, and the viewing distance, from CRT to the eyes of observers, was 500 mm. The stimuli to be measured were evenly distributed along 12 directions every 30' in (a*, b*)-plane and in 8 directions every 45' in (a*, L*)- and (b*, L*)-plane from the CIE Gray and Blue color centers,2,9) of which the CIELAB values are listed in Table 1. The CIE Gray and Blue centers were selected as the test color centers, be- cause gray is considered the most basic color and blue the most difiicult color to visually predict by color-difller- ence formulae, due to the most different perceptual char- acteristics in the blue region with respect to the tolerance ellipses not pointing toward the neutral point.5) The test stimulus was a square array of four I ' squares 142