A perceptual measure to predict the visual distinction between two color images

This paper presents a perceptual model to predict the visual distinction between a pair of color images which correlates well with the human observer's evaluation. Its basic assumptions are to be in agreement with spatial-frequency channels models quite successful for the detection of visual patterns. The model was applied to predict the perceived quality of a basic pattern as well as the visibility of the artifacts in JPEG reconstructions. The rank order produced by the PID error and other quantitative measure are compared with the reference rank order as given by a subjective rating performed by photointerpreters. keyword and phrases: visual distinction, JPEG compression, artifacts visibility, visual quality, opponent-colors space, log Gabor functions, xation points, integral features, sum-of-outputs rule.