A Color-texture Approach Based on Mutual Information for Multispectral Image Classification

Abstract

In this work we propose an approach to improve the results of color texture image classification. We construct a new space called hybrid color-texture space by selecting the most discriminating attributes for the textures. Attributes are calculating from the co-occurrence matrix. The selection is done by the algorithm MRMR based on the mutual information. The Support Vectors Machine classifier (SVM)is used. A comparison with an iterative selection is also performed. The effectiveness of the proposed approach is evaluated on the VisTex database and on a SPOT HRV (XS) image representing two forest areas in the region of Rabat.