Bayesian model of human color constancy

Abstract

Vision is difficult because images are ambiguous about the structure of the world. For object color, the ambiguity arises because the same object reflects a different spectrum to the eye under different illuminations. Human vision typically does a good job of resolving this ambiguity-an ability known as color constancy. The past 20 years have seen an explosion of work on color constancy, with advances in both experimental methods and computational algorithms. Here, we connect these two lines of research by developing a quantitative model of human color constancy. The model includes an explicit link between psychophysical data and illuminant estimates obtained via a Bayesian algorithm. The model is fit to the data through a parameterization of the prior distribution of illuminant spectral properties. The fit to the data is good, and the derived prior provides a succinct description of human performance.

Computation and visualization of the MacAdam limits for any lightness, hue angle, and light source

“…12 into account, where the shape and volume of the color solids are compared as a function of the degree of chromatic adaptation, it is clear that if our visual system did not use chromatic adaptation the number of discernible colors would be much greater, with minor perceptual correspondence among illuminants and light sources. Therefore, the color constancy phenomenon, [27][28][29][30][31] based at the first stages on a chromatic adaptation transform, means from a evolutionary point of view an adaptive mechanism to reduce the variability of perceptually noncorresponding distinguishable colors. Thanks to this adaptive mechanism, the result of the evolution of several million years, our primate predecessors achieved a quasi-invariant system for encoding color in front of chromaticity changes of ambient light, clearly advantageous for establishing quasi-constant recognition patterns of objects and scenes very important for survival.…”

Video-Based Illumination Estimation

“…A generalization of these two methods, called Shades of Gray [3], combines them using the Minkowski norm function. Other approaches based on more complex statistics of the image chromaticity distributions include the neural network approach [4], color by correlation [5], and the Bayesian approach [6]. All these algorithms are based on the zero-order structure of images, which means using the statistical distributions of the original pixel values to estimate the light source color of the image.…”

A spiking neural network for illuminant-invariant colour discrimination

“…The human visual system has the remarkable ability of perceiving and discriminating object colours largely invariant to changes in illumination [8]. However, attempts to mimic the human visual system invariance to the illuminant have not been yet successful [16], [21].…”