The RIT-DuPont visual color-difference data as the RIT-DuPont data set, and they are the basis for [Color Res. Appl. 16, 297-316 (1991)] have been used our current analyses. to estimate contours of equal color-differences (ellip-
The RIT-DuPont data set consists of 156 tolerances, soids) at 19 color centers, in CIELAB and x, y, Y/100 each of equal visual color difference, DV. Nineteen color color spaces. The ellipsoid fits are better in the CIELAB centers in CIELAB were defined sampling the gamut of space than in x, y, Y/100, since the design of the RITsurface-colors. At each color-center, samples were pre-DuPont experiment emphasized directional balance in pared varying in defined CIELAB directions. The samples CIELAB. The ellipsoids estimated are hardly tilted with were an acrylic-lacquer automotive coating sprayed onto respect to L* or Y/100, and they appear to be in overall primed aluminum panels and cut into 5 1 6.5 cm rectangood agreement with those reported for object colors in gles. In a given direction, samples were selected forming recent publications. From the characteristics and accu-5-7 color-difference pairs of varying magnitude about racy of the RIT-DuPont experiment, the current ellipthe color center. Principal component analysis was used soids can be considered highly reliable and representative to define the vector direction and actual color center for of color discrimination under the observational condieach set of color-difference pairs. Each pair was mounted tions employed, these closely following the ''reference edge to edge on gray panels with L* of 38. The directions conditions'' recently suggested by the CIE for industrial included each axis, interactions in chromaticness, and incolor-difference evaluation [Color Res. Appl. 20, 399-teractions in all three axes (defined as vector directions 403 (1995)].