Abstract
It is known that natural scenes provide certain statistical properties of the luminance distribution of surface colors in those scenes, which might be utilized by the visual system to achieve color constancy in the scenes. Uchikawa et al. (JOSAA, 2012) showed that the optimal-color luminance shell was similar in shape to the luminance distribution of surface colors in natural scenes, and proposed that the peak of an optimal-color luminance shell may be used to estimate an illuminant color in a scene. In this study we tested whether this optimal-color hypothesis was effective in non-natural scenes, which had biased-luminance distributions of surface colors. In experiments, the stimulus consisted of 61 hexagonal elements: a center test and 60 surrounding colors. Each element subtended 2-deg in diagonal. We employed three biased-luminance distributions of surrounding colors. They were made to have reduced luminances only in the red, in the green and in the blue region, respectively, from the optimal-color (reference) luminances. The mean luminance and chromaticity of surrounding colors were kept constant for all luminance distributions. The test illuminants were set at 3000K, 6500K and 20000K. The observer adjusted both luminance and chromaticity of the test element until it appeared as a white surface under a test illuminant. The results showed that fairly good illuminant estimations were obtained for all biased-luminance distributions, indicating that the visual system held color constancy in non-natural scenes. The optimal-color hypothesis can explain these illuminant estimations. We discuss effectiveness of the optimal-color hypothesis comparing with other hypotheses.