Comparison of several sparse reconstruction algorithms in SAR imaging

Abstract

As the effective means for achieving sparse microwave imaging, sparse reconstruction algorithms (SRAs) can be generally classified into four main categories: greedy pursuits, l1-norm minimization, nonconvex optimization and Bayesian framework. In this paper, we compare the performance of the typical SRAs in synthetic aperture radar (SAR) imaging. We consider four algorithms including the orthogonal matching pursuit (OMP), the iterative shrinkage-thresholding algorithm (IST), the iterative half thresholding algorithm (IHalfT) and the complex approximate message passing algorithm (CAMP), which are chosen from the aforementioned main categories respectively. In the light of theoretical analysis, we discuss the potential advantages of those algorithms applied to SAR, such as range-azimuth decoupling based 2-D reconstruction, adaptability to various modes of SAR imaging and parallel acceleration capability. On the basis of the results in the simulations and real data processing, the performance comparison of those algorithms is summarized in the end.