SAR image filtering based on the heavy-tailed rayleigh model

We describe a novel adaptive despeckling filter for Synthetic Aperture Radar (SAR) images. In the proposed approach, the Radar Cross Section (RCS) is estimated using a maximum a posteriori (MAP) criterion. We first employ a logarithmic transformation to change the multiplicative speckle into additive noise. We model the RCS using the heavy-tailed Rayleigh distribution, which was recently proposed as an accurate model for amplitude SAR images. We estimate model parameters from noisy observations by applying the 'method-of-log-cumulants', which relies on the Mellin transform. Finally, we compare our proposed algorithm with the classical Lee filtering technique applied on an aerial image and we quantify the performance improvement.