Abstract
The present study investigated whether and how the visual system computes a single color (global color) as a statistical summary representation of the individual element colors in multi-colored textures. The texture pattern (2.5 × 2.5°, 30 cd/m2) consisted of equiluminant small squares of 3.3 arc min and their chromatic variation roughly belonged to the same color category. Element colors in the texture were selected from a circular color distribution on the CIE u'v' chromaticity diagram. The color at the center of the distribution was one of four colors (orange, green, blue, and purple) which were located at the distance of 0.05 units from a white point (CIE D65). For each center color, 8 satellite and 8 intermediate colors were determined at the distance of 0.02 and 0.01 units from the center color, respectively. The relative number of the center-color elements to the satellite- and intermediate-color elements was systematically manipulated in the texture. The observer matched the global color of the multi-colored texture by adjusting the chromaticity of a spatially-uniform matching stimulus. Results showed that when all element colors were equally frequent, the global color deviated from the colorimetric average toward the color of the highest saturation. But when the relative number of the center-color elements became larger, the global color became closer to the colorimetric average. Similar results were found even when the texture was composed of only satellite and intermediate colors and the intermediate-color elements were more frequent. The influence of the highest saturation was quantitatively different for different color categories, but this can be accounted for if differences in cone-opponent activities were taken into account. These findings are consistent with the interpretation that the global color is determined as a weighted average of cone-opponent activities in which more salient elements (saturated or frequent) are given greater weights.
Meeting abstract presented at VSS 2014