A novel hierarchical bayesian method for SAR image reconstruction

Synthetic Aperture Radar (SAR) imaging involves an ill-posed inverse problem of reconstructing an image of an unknown scene (target) from partial and truncated information of its Fourier Transform (FT). Performances of conventional deterministic SAR imaging methods based on Inverse Fourier Transform (IFT) are limited by a fundamental assumption that unmeasured data in the Fourier domain are treated as zero. Indeed, these methods do not account for prior knowledge of the scene. Bayesian methods with appropriately chosen priors emerge as promising alternatives. We develop a hierarchical Bayesian method for SAR image reconstruction with a generalized Total Variation (TV) prior. We adopt a coordinate-descent optimization method for the MAP estimation. Compared to existing quadratic constraints (equivalent to a Gaussian prior) the proposed method with the TV prior has the capability of enhancing the region smoothness and preserving the edges between regions in the reconstructed SAR image. This paper demonstrat...