Color in Computer Vision: Fundamentals and Applications

Chapter 8: Illuminant Estimation and Chromatic Adaptation

Explicitly correcting an image for the color of the light source, producing a transformed version of the input image, is called color constancy. Human vision has the natural tendency to correct for the effects of the color of the light source, [108–111], but the mechanism involved with this ability is not yet fully understood. Early work by Land and McCann [13, 14, 112] resulted in the retinex theory. This theory posited that both the retina and the cortex are involved in the processing. Many computational models are derived on the basis of this perceptual theory, [113–115]. However, computational models can still not fully explain the observed color constancy of human observers. Kraft and Brainard [116] tested the ability of several computational theories to account for human color constancy, but found that each theory leaves considerable residual constancy. In other words, without the specific cues corresponding to the computational models, humans are still, to some extent, color constant [116]. Alternatively, observations on human color constancy cannot be readily applied to computational models either: Golz and Macleod [117, 118] showed that chromatic scene statistics influence the accuracy of human color constancy, but when mapped to computational models, the influence was found to be very weak at best [119]. Recent advances that are not pursued further in this book include the suggestion from computational color constancy ...