Mechanisms underlying segmentation of colored textures.

Abstract

This paper examines the nature of the post-receptoral chromatic and achromatic mechanisms involved in a texture segmentation task. Observers viewed a 64 x 64 square-element texture in which the chromaticity and brightness of each element was drawn from a one-dimensional Gaussian distribution of values centered at a white point in color space. The orientation in color space and variance of this distribution defined the noise in the stimulus. The mean chromaticity and/or brightness of a central 32 x 32 element area (the target) was shifted away from the white point along the same direction as the noise or along a different direction. We measured target thresholds as a function of noise amplitude. The steepness of this relationship defined the effectiveness of the noise. Within planes spanned by the achromatic axis and one of the two cardinal chromatic axes, we found selective effects of noise along each of the two cardinal axes. Within the purely chromatic (isoluminant) plane, we found selective effects of noise along four lines--the two cardinal chromatic axes and two intermediate ones. We describe a simple model to account for our results.