A Novel Iterative Thresholding Algorithm for Complex Image Based Sparse SAR Imaging

Abstract:
Compared with traditional matched filtering (MF)-based imaging method, sparse synthetic aperture radar (SAR) imaging, a combination of sparse signal processing and SAR imaging techniques, has shown significant potential to improve image quality. However, being stuck with the computational cost, it is difficult to apply conventional observation matrix based sparse SAR imaging method for practical large-scale scene reconstruction. In this paper, a complex image based sparse SAR imaging model is presented to reduce the computational complexity to the same order as that of MF, and simultaneously achieve a similar improved image performance as raw data based technique. In addition, a novel iterative thresholding algorithm, named as BiIST, is proposed and successfully used for the sparse reconstruction of considered scene. Compared to other regularization recovery algorithms, BiIST can obtain not only a sparse image of interested area, but also a non-sparse estimation of surveillance region with the same image phase information as and similar background statistical distribution to those obtained by MF-based method. Experimental results based on simulated and TerraSAR-X data validate the proposed method.