Alternating minimization algorithm for shifted speckle reduction variational model

Abstract

In synthetic aperture radar (SAR), the observed image is corrupted by the speckle (multiplicative noise). The variational models with the total variation (TV) regularization have attracted much interest in reducing the speckle due to the edge preserving feature of TV. Recently, several TV regularized convex variational models, such as the maximum a posteriori (MAP) model for a log-transformed image and the I-divergence model, have been proposed. In this paper, we adapt Tseng's alternating minimization algorithm to solve the proposed shifted speckle reduction variational models with TV. The algorithm for the proposed shifted variational models does not require any inner iteration or inversion involving the Laplacian operator that is required in recent algorithms based on an augmented Lagrangian framework. Hence, the proposed method is very simple and highly parallelizable and so efficient to despeckle huge SAR images.