Towards a Model of Color Reproduction Difference

Abstract

It is difficult to predict the visual difference between cross-media color reproductions. Typically, visual difference occurs due to the limitations of each output medium's color gamut, the difference in substrate colors, and the gamut mapping operations used to transform the source material. However, for pictorial images the magnitude of the resulting visual difference is also somewhat content dependent. Previously, we created an interval scale of overall visual difference ($\Delta V$) by comparing gamut mapped images side-by-side on a variety of simulated output media. In this paper we use the preexisting visual difference data, together with the known source images, as well as information relating to the output gamuts, to create a model of color reproduction difference which is both output-gamut and source-image dependent. The model generalizes well for a range of images, and therefore performs better than mean $\Delta E_{00}$ as a predictor of visual difference. In addition, the inclusion of coefficients derived directly from the source images provides insight into the main drivers of the visual difference.