Abstract
Color discrimination in dichromats is impaired but the extent to which their color constancy is affected remains unclear. This work compares color constancy for red-green dichromats with normal observers. Stimuli were generated using spectral reflectance data of natural scenes obtained from hyperspectral imaging. Two scenes of rural and two of urban environments were selected from the database of Foster et al. (2006, J. Opt. Soc. Am. A, 23, 2359). Each scene was rendered with correlated color temperature (CCT) in the range 4012-40231K (steps of 23.3 MK-1) along the daylight locus. In this condition of pure illuminant change average luminance was 10 cd/m2. Thresholds for detecting a change in CCT were estimated with a 2AFC experiment with the stimuli presented on a calibrated CRT controlled by a ViSaGe MKII (Cambridge Research Systems) in 24-bits-per-pixel true-color mode. Similarly to Pearce et al. (2014 PLoS ONE 9(2): e87989), observers viewed the scene illuminated by 6700K and then two images, one illuminated by either a higher or lower CCT. The observers had to identify the image that looked different. Thresholds for a pure luminance change were also estimated with similar methodology but varying only the average luminance of the scenes. Four normal observers, two protanopes and one deuteranope were tested. Four mixed-model ANOVA analyses didn’t reveal any significant main effect of scene or of its interactions with group of observers (all P > 0.05). However, a group effect was found for changes towards lower CCT [F(2,4) = 8.29, P < 0.05] but indicating slightly higher color constancy for dichromats. These results suggest that for red-green dichromats color constancy mechanisms along daylight locus are at least as efficient as for normal observers.
Meeting abstract presented at VSS 2015