Orthogonal Matching Pursuit Algorithms based on Double Selection Strategy

The greedy algorithm is a promising signal reconstruction technique in compressed sensing theory. The generalized orthogonal matching pursuit (gOMP) algorithm is widely known for its high reconstruction probability in recovering sparse signals from compressed measurements. In this paper, we introduce two algorithms based on the gOMP to address the signal reconstruction issue. In these two approaches, the double selection strategy is exploited to automatically select a more suitable reconstruction method according to the change of the support set. Therefore, the proposed methods have greater flexibility in atom selection and also can remove the erroneous atoms in the support set to enhance the reconstruction accuracy when compared to the gOMP. Simulation results show that the presented algorithms have better recovery performance for both one-dimensional sparse signals and two-dimensional image signals.