Abstract
The aim of this study was to estimate discrimination thresholds in 3-D colour space for images of skin patches from two ethnicities - Caucasian and Asian; with the observer fully immersed in three different illumination conditions - dark, daylight and cool-white-fluorescent. In the illuminated conditions, the patches were presented in an ‘object mode’ wherein the screen was covered by grey cardboard and only the stimulus was visible. Thresholds were measured using a 4-AFC paradigm. The stimuli were generated by adding a 3-D test colour vector to the original skin patch. The test vector was in one of 14 pre-selected directions in CIELAB colour space, and its length was controlled by the QUEST adaptive procedure. The final length of the vector in each direction was the estimated threshold. An ellipsoid was fitted to the 14 data points in the 3-D CIE UCS (u’v’Y) space. For each observer, this fit was done for each illumination condition and patch type separately (six fits per observer). The lengths of the three axes and the orientation of each ellipsoid were calculated. Here we report data for 5 observers. Our main findings are: (1) Discrimination ellipsoids for D65 (daylight) and TL84 (cool white fluorescent) are similar in size and orientation (2) One of the axes of the ellipsoids is always parallel to the luminance axis (3) In the u’-v’ plane the projected ellipses are oriented at an angle i.e. neither of the axes is parallel to the u’ or v’ axes (4) Variability of the thresholds was larger in the dark condition. Our preliminary results (five observers) suggest that tolerance limits for skin colour changes are very similar for daylight and fluorescent office light. Discrimination along the luminance axis (Y) seems to be independent of discrimination in the chromaticity plane (u’v’).
Meeting abstract presented at VSS 2015