Fully polsar image classification using a diffusion-reaction system enhanced with morphological filtering

Abstract

Accurate image classification of Fully PolSAR (Polarimetric Synthetic Aperture Radar) images is of maximum interest in monitoring remote sensing data. Such classification is a difficult task because of the presence of speckle. Diffusion-reaction systems (DRS) using stochastic distances have been recently applied to successfully despeckling and classifying PolSAR images. In this work we show that better results are obtained by including an iterative morphological schema coupled to the DRS. Experimental results illustrate that the proposed approach significantly improves the visual quality and classification measures of quality of the outcomes.