Towards Adjacency-Aware Categorical Color Palettes: Algorithm, Evaluation, and User Preferences

Abstract

As data complexity increases, so does the need for effective color encoding in visualizations. Existing palette generation tools optimize global discriminability but overlook adjacent luminance contrast, limiting legibility in stacked bar charts, pie charts, and similar visualizations. As a first step toward adjacency-aware palette optimization, we extend an established categorical palette generation algorithm with an adjacent luminance contrast term, integrated alongside color distance and name difference via Simulated Annealing and aligned with WCAG accessibility guidelines. A benchmark evaluation against two established optimizations shows that incorporating adjacent contrast maintains comparable discriminability while significantly improving WCAG contrast adherence. An exploratory preference study provides initial evidence that resulting palettes remain aesthetically competitive. Our findings indicate that adjacent luminance contrast can be jointly optimized with global color discriminability without meaningful trade-offs, laying the groundwork for studying its impact on chart reading performance and exploring more sophisticated aesthetics optimization.