Abstract
We previously developed the optimal color hypothesis to estimate an illuminant in a scene. It assumes that the visual system has a priori knowledge about optimal colors under different illuminants and selects the optimal color luminance-shell that gives the best-fit to the luminance distribution of chromatic stimuli in a scene. The peak of the selected optimal color luminance-shell corresponds to the illuminant color in the scene. This hypothesis has well explained color constancy in various stimulus conditions. This study aimed to verify the validity of the hypothesis using luminosity thresholds for chromatic stimuli. The thresholds were obtained in our previous experiments, where an observer adjusted the luminance of a test stimulus until it appeared to be brighter than the upper limit of the surface mode and dimmer than the lower limit of the luminosity mode (luminosity threshold). A circular test stimulus was presented at the center of surrounding stimuli on a CRT monitor. We selected 15 reflectances for the test stimulus and 180 reflectances for the surrounding stimuli from Brown's 574 spectral reflectance data of natural objects. We employed three luminance-chromaticity distributions (normal, flat, and valley) of the surrounding stimuli and three color temperatures (3000K, 6500K, and 20000K) of the illuminant. It was shown that the optimal color hypothesis well predicted the luminosity threshold function of the test stimulus chromaticity in terms of its shape and the peak regardless of the luminance-chromaticity distributions of surrounding stimuli and color temperatures of the illuminant.