Speckle Suppression in Multi-Channel Coherent Imaging: A Tractable Bayesian Approach

Abstract

Coherent imaging methods like synthetic aperture radar (SAR) are subject to speckle, and the suppression of this noise-like quality is often considered a prerequisite to image interpretation. Likewise, circumstances such as frequency jamming or system multiplexing produce gaps in the data and corresponding image artifacts, which further impact the utility of resulting imagery and induce correlation in the speckle process. For multi-channel SAR configurations such as the polarimetric or interferometric modes, a Bayesian imaging procedure is proposed to harvest the benefits of jointly processing channels in order to recover scattering information. The inference technique recovers the per-pixel multi-channel SAR covariance and incorporates a statistical model of speckle and a priori knowledge of the varieties of clutter present in the scene. Further, by combining speckle reduction into the image formation process for gapped aperture measurements, the estimator can balance competing objectives of image resolution, side-lobe suppression, texture preservation, and speckle reduction. An expectation-maximization algorithm is made computationally tractable by a graph-coloring probing technique to provide the block-diagonal portion of a large matrix inverse. Recovery results with simulated and measured fully polarimetric airborne SAR data indicate that the proposed method reduces the appearance of speckle and preserves spatial resolution.