Classification of Polarimetric SAR Image using JS-Divergence Profile

The effectiveness of a classifier is dependent on how accurately it classify the boundary samples. To enhance the discrimination between the boundary pixels, in this research we propose a novel feature extraction technique that generates a profile of the polarimetric synthetic aperture radar (PolSAR)image by using Jensen Shannon Divergence (JSD). The experiment conducted on two real PolSAR data sets using support vector machines (SVM) classifier shows the potentiality of the proposed profile. For both the data sets, the proposed profile provides at least 3% improvement in the classification accuracy as compared to the traditional T3db features.