Texture comparison based on selected texture primitives

Abstract

The texton theory proposed by Julesz defines textons as hypothetical basic elements for texture images. In this research, we propose an operational interpretation of textons and show experimentally that they can be used in a similarity measure for texture discrimination and characterization. The method is based on evaluation of the complete set of primitives to obtain a smaller set for texture analysis. We call these texture primitives. Similarity between pairs of textured images is defined as a sum of the similarities of the individual selected texture primitives in both images. Extensive experimentation in texture characterization and discrimination using prototype texture images were performed. The results show that the method achieves a large amount of representation reduction (>90%) and yet the proposed similarity obtains impressive results in texture characterization and discrimination.