PolSAR Speckle Filtering Techniques and Their Effects On classification

In this paper the most significant and recent polarimetric SAR (PolSAR) speckle filters that have been developed so far are reviewed and the effects of them on classification and their ability to preserve polarimetric properties along with their strength in noise reduction are studied. For this purpose, the PolSAR filters such as refined Lee, intensity driven adaptive neighborhood (IDAN), scattering model base (SMB), speckle reduction anisotropic diffusion (SRAD), and a Non-Local PolSAR filtering are tested. These filters have been applied to a C-band multilook Polarimetric data taken by Radarsat-2 sensor over San Francisco Bay area, then the performance of the filters have been evaluated from various aspects: (1) land-use classification (2) speckle reduction level (3) edge preservation (4) polarimetric information maintenance.