A new nonlocal TV-based variational model for SAR image despeckling based on the G0 distribution

In this paper, we propose a new variational model for speckle reduction of synthetic aperture radar (SAR) images based on the G0 statistical distribution and nonlocal total variation (NLTV) regularization. The existing variational models for SAR despeckling regard the terrain backscatters as nonrandom, which is only suitable to depict the homogeneous regions. For inhomogeneous scenes, the backscatter fluctuations should be taken into account. Motivated by this, the inverse Gamma distribution is used to model the statistical property of the underlying terrain backscatter. By using the maximum a posteriori rule and NLTV penalty, a new variational model, named G0NLTV, is derived. The parameters in the model are estimated based on the Mellin transform of the G0 distributions according to the local heterogeneity. Since this model lacks the global convexity, a convex model is obtained by utilizing the logarithm transformation and variable substitution. Then, the primal-dual algorithm is used to solve this optimization problem. Experimental results on both synthetic and real SAR images demonstrate the effectiveness of the proposed method.