High contrast color sets under multiple illuminants

Abstract

In the modern digital society colors are often used to encode information. Nevertheless the selection of set of colors that maximizes class discriminability for nominal coding is a non-trivial problem. In this work we compare four different heuristics for the selection of sets of colors with fixed cardinality and maximum dissimilarity. The performance of each algorithm is evaluated both on single and multiple illuminants, on a sample of 1268 colors from the Munsell atlas, using ΔE 76 euclidean metrics on the perceptually uniform CIE L* a* b* space. Results are presented for color sets with cardinality up to 25. © 2013 Springer-Verlag.