Fast Orthogonal Matching Pursuit Reconstruction Algorithm Based on Particle Swarm Optimization

Abstract

Utilizing orthogonal matching pursuit (OMP) algorithm could achieve the compressed sensing (CS) reconstruction of hyperspectral image (HSI). Due to the large number of bands, high computational complexity of OMP could not meet real-time processing requirements. Aiming at this defect, a fast OMP reconstruction algorithm based on particle swarm optimization (PSO) is proposed. PSO algorithm which is more powerful in searching for local optimal solution is applied to optimize matching process of OMP. In addition, Hermitian inversion lemma is explored to update the residual in an iterative way, leading to further improve the efficiency. Experimental results demonstrate that compared to OMP, the computational efficiency of proposed algorithm could be increased by 18 times while keeping high reconstruction accuracy.