Polarimetric Image Classification using Texture Features based on the Gray Level Co-occurrence Matrix and Local Binary Patterns

The paper assesses the use of two of the most common texture information extraction techniques for the classification of polarimetric Synthetic Aperture Radar (PolSAR) images and proposes a traditional machine learning framework for the classification step, using a fused feature vector. We consider both features extracted from the Gray Level Co-occurrence Matrix (GLCM) and features extracted using Local Binary Patterns (LBP) - based approaches. The feature vector integrating both types of information is computed on a pre-smoothed version of the input data and classification is done using Support Vector Machines (SVM). We evaluate our approach on synthetic data and on real polarimetric images. The experimental results show that the proposed solution provides good accuracy figures for high resolution images even when the number of training samples is low.