Speckle reducing in PolSAR images for topographic feature extraction

Abstract

This article presents a new approach that reduces the speckle in PolSAR (Polarimetric Synthetic Aperture Radar) image, and then improves the visualisation and facilitates operators to extract topographic planimetric features. The proposed filter takes into consideration all channels simultaneously (polarisation modes: HH, HV, VH and VV) for each channel despeckling process. So, for a given pixel in an image channel, the adjustment degree of its radiometric value will be decided by its adjacent pixels in the same channel and their equivalent pixels in other channels. Moreover, to avoid error propagation between channels, the filter runs in several iterations modifying the less ambiguous pixels first and leaves the rest of the pixels for a possible adjustment in next iterations. To modulate the uncertainty and combine inter and intra channel information and make decision, the proposed filter calls rules and techniques of Dempster–Shafer theory. The experimentation is based on Radarsat-2 images and our study areas were in Quebec City and the Arctic in Canada. The results of our experimentation show clearly the performance and the accuracy of our approach in spite of its simplicity and processing speed compared to other approaches.