Fusion Forward-Backward Pursuit Algorithm for Compressed Sensing

摘要

The Forward-Backward Pursuit (FBP), which is a recently proposed method, receives wide attention due to the high reconstruction accuracy. In this paper, we use the fusion strategy and propose the Fusion Forward-Backward Pursuit (FFBP) algorithm. This strategy only needs the reconstruction information of two FBP with different parameters. According to the termination conditions of the FBP algorithm, FFBP adopts different operation strategies, during the signal reconstruction. Without other priori information, FFBP effectively improves the exact reconstruction rate, compared with the original algorithm. Moreover, FFBP, which fuses two FBP with non-optimal parameter, can reconstruct a better signal than a single FBP with optimal parameter. We demonstrate the advantage of the proposed method through numerical simulations.