Finding a perceptual uniform color space with evolution strategies

One of the goals of color science is to find a color space in which the Euclidean distance between any two colors corresponds to the color difference perceived by humans. Various empirical measurement methods are known in order to obtain data about the perceived differences between selected color samples. The problem is to find a function which transforms a well-known color space into a new one which matches the empirical data as closely as possible. One approach is to use multidimensional scaling in conjunction with tri-linear interpolation. In this paper, a new method is presented, using an evolution strategy for finding control point positions of a free-form deformation. Objective functions that are very well suited for two kinds of empirical data are described. In various test cases, this method proved to be very reliable in finding good solutions. Furthermore, free-form deformation has some essential advantages against tri-linear interpolation, e.g. derivative continuity to any degree.