Image Classification Using Texture Features and Support Vector Machine (SVM)

Abstract

Due to their efficiency, texture features are frequently used for describing visual content of images. In this paper, we compare six widely used texture features namely, Weber Local Descriptor (WLD), Local Binary Pattern (LBP), Gist and Gray- Level Co-occurrence Matrix (GLCM), in addition to two recent ones namely, Three-Dimensional Connectivity Index (TDCI) and Dense Micro-block Difference (DMD). Moreover, we have proposed an improvement of TDCI so it can capture local variation of motifs instead of the global. As a classifier, we have considered using Support vector Machine (SVM). After conducting a detailed evaluation on four well-known texture benchmarks which are Broadatz, Vistext, Outext and DTD, we have found out that WLD has, in average, the best performance compared to the other features.