Pseudo-granulometry and morphological covariance for color psoriasis image segmentation

Texture features are very popular and widely-used in many image processing and computer vision applications. Mathematical morphology offers a series of tools for texture description, like granulometry and morphological covariance. However, there are few methods for color texture description through morphological approaches, due to the issues arisen at the extension of the morphological operations to multivariate images. In this paper we derive the pseudo-granulometry and morphological covariance through a recent approach of multivariate mathematical morphology.We also discuss two issues arisen when applying these measures to color images i.e. the pertinence of the morphological opening operation for color images and the computation of the color image volume. We apply the proposed measures using a k-means classifier for the segmentation of color images of psoriasis lesions. We present our results and perform a comparison with other approaches.