Unsupervised colour coding for visualization of categorical maps

Abstract

ABSTRACTAn unsupervised algorithm for colour coding of categorical maps is proposed. Our approach is aimed at maximizing colour differentiation between spatially adjacent categories, given a colour palette. The algorithm is specifically designed for maps characterized by the presence of numerous small regions of different categories. The proposed methods relies on a category co-occurrence matrix evaluated locally within non-overlapping and contiguous square blocks of sizes approximately equal to the 2 CIE (Commission Internationale de l’Éclairage) standard observer. The algorithm determine the colour-class association that maximizes the average of the minimum colour contrasts evaluated locally within each block. The algorithm is formulated as an optimization problem on a set of colour permutations. Examples are provided illustrating the performance of the proposed method.