Evaluating the Perceptual Uniformity of Color Sequences for Feature Discrimination

Abstract

Probably the most common method for visualizing univariate data maps is through pseudocoloring and one of the most commonly cited requirements of a good colormap is that it be perceptually uniform. This means that differences between adjacent colors in the sequence be equally distinct. The practical value of uniformity is for features in the data to be equally distinctive no matter where they lie in the colormap, but there are reasons for thinking that uniformity in terms of feature detection may not be achieved by current methods which are based on the use of uniform color spaces. In this paper we provide a new method for directly evaluating colormaps in terms of their capacity for feature resolution. We apply the method in a study using Amazon Mechanical Turk to evaluate seven colormaps. Among other findings the results show that two new double ended sequences have the highest discriminative power and good uniformity. Ways in which the technique can be applied include the design of colormaps for uniformity, and a method for evaluating colormaps through feature discrimination curves for differently sized features.