*** A Bayesian Approach to Color Constancy

I have developed a database of 1500 images of 100 objects under 5 different lighting conditions and in front of 3 different backgrounds (link). I've used this database as well as a number of currently published color constancy algorithms to compare and contrast the efficacy of given assumptions that are typically made in the resolution of the basic color constancy problem of assessing and correcting for unknown illumination. Sample images of objects as well as the effects of running a simple grey world algorithm on these same images can be seen here. The grey world algorithm works by simply removing the average color of the image from the total image under the assumption that any color deviating from white in the image comes from the lighting itself. Overall brightness is preserved.