A new medium-to-large colour-difference data set (mean DeltaE* = 8.25) usin
g textile samples and a gray-scale method has been produced with the aim of
comparing currently available datasets, putting particular emphasis on the
characteristics of lightness and hue dependencies of major advanced CIELAB
-based colour-difference formulae. The levels of errors in instrumental col
our measurements and in visual assessments were similar to those of previou
s workers. The random and large scattering nature of the trend of unit ligh
tness tolerances with lightness (L*) has made it difficult to apple, a nonl
inear lightness weighting calculation, e.g., that of CMC(l:c). The chroma p
osition corrected hue tolerances have shown a weak trend with hue-angle (h(
ab)) and it can be incorporated in the existing modified CIELAB formula, e.
g., CIE94, to improve its performance. Among the six formulae (CIELAB, CMC,
CIE94, LCD, BFD-II, and DCI-95) tested, the LCD (Leeds Colour Difference)
formula that is a modified model of the CIE94 colour-difference equation ha
s given the best performance. More visual experimental results are still th
ought to be required for further development in industrial colour-differenc
e evaluation. (C) 2001 John Wiley & Sons, Inc.