TomoSAR imaging based on random support split matching pursuit algorithm

In this paper we propose a new matching pursuit algorithm, termed Random Support Split Matching Pursuit (RSSMP), to solve tomoSAR imaging problem in compressed sensing form. This method pursues more than one atom at a time, randomly splits the pursued atoms to two subset, choose one subset with smaller residual energy, truncates this subset by descending modulus of amplitudes, and estimates sparse signal by least squares on the truncated subset. The advantage of our proposed algorithm is speed: using Orthogonal Matching Pursuit (OMP) algorithm as benchmark, RSSMP algorithm can save at least 50% run time for middle scale problem, 75% run time for large scale problem, and it has ability to save more run time by parallel computing. Experiments using simulation data and real world electromagnetic data validate the proposed method.