Data Mining and Its Use in Texture Analysis

Abstract

This paper deals with an idea of the use of data mining approach in texture analysis. A new method based on association rules is proposed. This method utilizes the multiresolution analysis of an analyzed image texture and works at the texture primitive level. Within this method, a technique for feature vector construction without any a priori knowledge of textures different from the analyzed one is presented. The contribution also contains some results of texture image segmentation experiments.