Visualization of Multichannel Images Using Differences in Human Perception of Brightness and Chromaticity

The work studies multichannel image visualization that is dimensionality reduction preserving perceptually critical information with inherent data loss. We investigate previously proposed Sokolov’s method, which leverages human perceptual differences between brightness and chromaticity. The assumption is the method yields maximally visually informative representation while minimizing perceptibility of visualization errors. To test this, we conducted a human study comparing Sokolov’s method against PCA and UMAP. Survey results from 62 subjects demonstrate Sokolov’s significant superiority in preserving local contrast.