One-class texture classifier in the CCR feature space

Abstract

A one-class classifier in the coordinated clusters representation (CCR) feature space is proposed in order to form classes with a required homogeneity with respect to given features of texture images. Experiments with gray level texture images of Brodatz catalog and “Rosa Porriño” polished granites demonstrate high discrimination performance of the criterion. The compactness of a class of images in the CCR feature space is determined by the class variance and, in addition, controlled by an empirical adjustment parameter.