Melgosa Manuel – Spain, University of Granada
Most experiments leading to the development of color-difference formulas (e.g. CIEDE2000, currently recommended by CIE and ISO) were based on visual assessments of flat samples with homogeneous or slightly textured colors. As a more realistic approach to many practical situations, in connection with the goals of CIE TC 8-17 “Methods for evaluating color difference between 3D color objects”, this paper reports visual and instrumental color-difference results from an experiment using metallic automotive samples with a size of 10.0 cm x 14.7 cm, including solid and effect (gonioapparent) nearly achromatic colors. Specifically, we used samples with 3 different convex curvatures: Flat samples (F), and samples with cylindrical curvatures of 10.5 cm (C1) and 18.0 cm (C2). 13 color pairs with each of the 3 mentioned curvatures were visually assessed in random order by a panel of 15 observers with normal color vision in a multi-angle byko-spectra color assessment cabinet, using the gray scale method. Each color pair was assessed 3 times by each observer. The average visual differences (∆V) ± standard deviations were 2.89±1.39, 2.60±1.19 and 2.49±1.22 CIELAB units, for color pairs with F, C1 and C2 curvatures, respectively. Therefore, on the average, the visual differences ∆V decreased 0.4 CIELAB units for C2 with respect to F. We found statistically significant differences associated to curvature for visual differences in 4 of the 13 color pairs assessed, all of them with samples with effect colors. While samples in the 13 color pairs were selected from instrumental color measurements performed with a BYK-mac multi-angle spectrophotometer, the incidence and reflection angles in the multi-angle byko-spectra cabinet were considerably different than those employed by this spectrophotometer, particularly for samples with curvatures C1 and C2. Therefore, for each sample we performed additional instrumental measurements in 4 specific points, using a CS-2000 spectroradiometer placed at the same position than the observers. Best correlations between instrumental color-difference measurements (∆E) made by the BYK-mac instrument and the CS-2000 spectroradiometer were found when the spectroradiometer was focused at the center of the samples, with independence of curvature degree. Regarding the strength of the relationship between visual (∆V) and instrumental (∆E) measurements, from instrumental spectroradiometric measurements at the center of the samples, for F/C2 curvatures we found STRESS values (CIE 217:2016) of 54.4/48.7, 61.8/57.7, and 42.5/37.0 using the CIELAB, CIEDE2000 and AUDI2000 color-difference formulas, respectively. Beside big differences between performance of best (AUDI2000) and worst (CIEDE2000) formulas in this experiment, they were not statistically significant. More research is necessary to propose best practices for the use of conventional color-difference formulas in real situations involving 3D object colors.