A Novel Approach for Urban Unsupervised Segmentation Classification in SAR Polarimetry

Abstract

In this study, we propose a novel unsupervised classifier for polarimetric SAR image segmentation by using Kullback-Leibler (KL) Divergence and subsequently the K-means algorithm. The K-means algorithm is used to divide the SAR data into spatial clusters based on the mutual proximity of the data points, whereas the KL Divergence or relative entropy measures the difference between one probability distribution and another. The main aim of this study is to assess the performance of classical approaches with this novel approach and provide better results for more accurate SAR image classification.