Sparse Reconstruction Using Improved SL0 Algorithm by Tabu Search and Application to Spotlight SAR Imaging

Abstract

Recently, researchers have made a lot of effort to reduce the cost of Synthetic Aperture Radar (SAR) imaging with the Compressed Sensing (CS) theory, of which sparse reconstruction is an important part. One method for solving sparse reconstruction problem is Smoothed L0 (SL0) algorithm, in which the L0 norm is approximated with a convex function. By minimizing the convex function, the feasible region constraint is satisfied. In this paper, we present the Improved SL0 (ISL0) algorithm by changing the optimization stage of the SL0 and performing it as Tabu Search (TS) algorithm. By replacing the steepest descent algorithm with the tabu search algorithm, we achieve a faster iteration speed with higher recovery quality using ISL0 compared to the SL0. Regarding the results of numerical experiments under the same test conditions, the number of iterations for the new algorithm compared to the original SL0 was about 20 times less. The efficiency of the ISL0 is evaluated for the reconstruction of spotlight SAR images through an experiment. The results of this simulation indicate that the quality of image reconstruction with ISL0 is better than SL0 for various SNR.