Creating perceptually distinguishable categorical colormaps remains a major challenge
in dense multiclass visualizations. In these settings, neighbouring categories can be difficult to distinguish. Existing colourmap optimization approaches commonly rely on perceptual colour difference metrics, such as CIEDE2000 (ΔE). However, the extent to which numerical perceptual improvements yield measurable gains in human performance remains insufficiently understood in complex visualization contexts.
This PhD project studies context-aware colormap optimization, integrating perceptual colour difference with context aware metrics. This integration aims to improve colour discrimination in dense scatterplots. Three optimization algorithms are applied across multiple standard colormaps: Simulated Annealing, Genetic Algorithm, and
Particle Swarm Optimization. These algorithms maximize perceptual separation while accounting for neighbouring contextual relationships between categories. The framework is evaluated through quantitative analysis and controlled user studies involving discrimination and pairwise comparison tasks, with class counts of K = 5, 10, 20, and 30.
Initial results have shown that optimization generally improves perceptual separation according to the optimization objective across evaluated conditions. However, human performance outcomes are strongly influenced by the visualization context and task conditions. Improvements in perceptual optimization metrics do not consistently translate into better discrimination accuracy or perceived discriminability. These findings suggest that perceptual colour metrics alone may not fully capture graphical perception in dense multiclass visualizations.
Team
- Fatimah Alqahtani (PhD student), Newcastle University
- Dr Sara Johansson Fernstad (main supervisor), Newcastle University
- Dr Stephen McGough (co-supervisor), Newcastle University