Perceptual Colour Correlogram and Perception-based Statistical Features of Colour Texture

In this paper a novel definition and understanding of colour correlogram has been proposed. The proposed colour correlogram generalizes the spatial graylevel dependency matrix (SGLDM) to the case of colour textures. This generalization is based on perceptual colour difference measure expressed in the language of the CIELab colour space components. Application of the colour difference instead of arbitrary colour indices or colour components themselves allows to avoid colour-shuffling palletization and introduction of multidimensional objects, respectively; the proposed perceptual colour correlogram is a single 2D matrix. At the same time, a simple relation of the proposed colour correlogram to the spatial graylevel dependency matrix for graylevel textures is retained. Based on this relation it will be shown that there exists a vector of statistical features built from the perceptual colour correlogram which can be used to describe textures in perceptual terms. These statistical features and their abovementioned perceptual interpretation generalize Haralick concepts derived for the SGLDM.