MAP Filtering for SAR Images Based on Heavy-Tailed Rayleigh Modeling of Speckle

Traditional Rayleigh distribution cannot accord with the heavy-tailed statistics of speckle because of the use of central limit theorem. In this paper, speckle in synthetic aperture radar (SAR) amplitude image is modeled as heavy-tailed Rayleigh distribution based on the non-Gaussian assumption of complex echo in each resolution cell, and the maximum a posteriori (MAP) filter is presented using gamma prior distribution. Based on Mellin transform, parameters of heavy-tailed Rayleigh distribution are estimated from the observed image. The de-speckling experiments and their quantitative measures demonstrate that the MAP filter based on heavy-tailed Rayleigh modeling of speckle owns higher capability of noise suppression compared to the one using the traditional Rayleigh distribution and the linear minimum mean square error (MMSE) filter.