Distributed target SAR image de-blurring using phase gradient autofocus

Abstract

A new analysis of errors in blur function estimate formed by the real-time phase gradient autofocus (PGA) algorithm is presented for synthetic aperture radar images of distributed targets in correlated noise clutter. It is shown that the PGA algorithm, like the Attia-Steinberg and Vachon-Raney focusing algorithms, can estimate a translation-invariant blur function when no point reflectors are present. The analysis shows that simulation evaluations of PGA performance which do not include sufficient simulated clutter can tend to underestimate the performance of the algorithm in initial iterations. Implications of the error analysis for performance optimization of real-time PGA implementations are presented for the algorithm steps that involve range bin selection, circular shifting, and windowing. It is shown that range bins selected for processing should be widely spaced if possible. If distributed targets are present which cause locally spatially correlated imagery, then it is desirable that the circular shifting segment of the algorithm maintain maximum decorrelation of the intermediate windowed and aligned images used for iterative phase error estimation.