CNP2D: The Chromatic Noise Pattern Discrimination Dataset

Pixel-wise color difference metrics like ΔE00 have long been used in image analysis, but it remains unclear how scores should be integrated over space. To highlight this, a psychophysical experiment was conducted to characterize visual sensitivity to differences in chromatic noise patterns in different color and pattern contexts. The results demonstrated that observers were more sensitive to chromatic noise pattern (CNP) differences when similar colors were spatially dispersed over the pattern as opposed to clustered. Further analysis with common image color and texture difference metrics showed that none were sensitive to this effect. This finding highlights the need for metrics which capture the perceptual interaction between color and texture.