A comparative study on texture features used for segmentation of images rich in texture

Abstract

A comparative study based method to select appropriate texture feature for image segmentation using K-means clustering algorithm is proposed in this paper. We study and record the performances based on three features namely, Contourlet, Gabor and Tamura. An enhanced version of Tamura feature is proposed that produces better result than the conventional one. Results of our experiment suggest that, for a given class of images, segmentation algorithm using Contourlet and Gabor with similar feature space perform equally well. On the other hand, performance of conventional Tamura feature lacks consistency but Tamura with multiple coarseness and directions improves segmentation.