Computational models for color perception

Shigeki Nakauchi

The goal of this study is to understand functional role of color vision: coding, representation and interpretation of color information by the visual pathway. To this end, a computational models are developed to explain various phenomena relating to color perception, including color transparency, constancy, categorical color perception and visual attention to color.

Since the NRV project has started, color transparency have been studied as one of the target phenomena of the computational theory. Perception of transparency refers to the phenomenon that an observer sees not only surfaces behind a transparent medium but also the transparent medium itself. The proposed theory successfully decompose a given color image into multiple layers by determining the depth ordering of surfaces and the type of color mixture, according to the physical realizability of recovered surface properties.

The heart of computational theory/modeling of color perception is shedding light on a problem how color information is represented in the visual pathway. Several types of experiments provide evidence that imply the existence of higher-order color mechanisms which visual cortex is thought to be a more likely substrate for. Psychophysical experiments are now being planed to explore the color representation mediating color detection or discrimination or color segregation. Experimental results should provide hints for mathematical description of the color representation in the computational model of color perception.