Texture discrimination was the second more important task studied after colour perception and measurement. A lot of works have explored it using a separated channel processing and very few have addressed the vector aspect of this spatio-chromatic information. In this paper we propose a novel vector processing for colour texture characterization: the Colour Contrast Occurrence matrix (C_2O). The C_2O is expressed using a perceptual distance in the CIELab colour space and two angles characterizing the chromaticity, and darker or lighter direction of local differences. The set of local differences, the contrast occurrences, is represented in a 3D representation, offering an understandable representation of the texture variations. In this work, we analyze also the feature invariance to changes in illumination, viewpoint and spectrum of the light source. Performances in classification tasks on several texture databases show the added-value of the C_2O for texture discrimination especially when the texture content becomes complex.
展开▼
