Texture classification using third order correlation tools

Abstract

This study presents a new method for textures classification based on higher order statistics (HOS). We propose the use of third order correlation tools for texture analysis. We compare the performance of three different tools: the bicorrelation in the spatial domain, the bispectrum in the frequency domain and the bicorspectrum which is a spatial/frequency representation in that case. We test classification on representative textures of Brodatz album.