An segmentation-based PolSAR image classification method via texture features and color features

This paper describes a new segmentation-based classification technique for fully polarimetric synthetic aperture radar (PolSAR) images. Based on the framework which conjunctively uses statistical region merging (SRM) for segmentation and support vector machine (SVM) for classification, we improve the method by jointly introducing texture features and color features. For the segmentation step, to guarantee the regional homogeneity, we propose a new local binary pattern (LBP) operator, regional homogeneity local binary pattern (RHLBP). For the classification step, we introduce the color features to improve the classification accuracy. Texture features and color features can provide discriminative information to separate those pixels and regions with similar polarimetric features which are from different classes. Extensive comparison experimental results with state-of-the-art methods on the L-band PolSAR data have demonstrated the effectiveness of our proposed method for PolSAR image classification.