A neighborhood-based multiple orthogonal least square method for sparse signal recovery

Abstract

Orthogonal least squares (OLS) is a popular greedy algorithm because of its simplicity and low complexity. To improve the reconstruction accuracy performance of OLS, the multiple orthogonal least squares (MOLS) algorithm selects multiple L indices per iteration. Compared with OLS, MOLS improves the reconstruction accuracy and reduces the running time. However, the MOLS algorithm ignores the situation that the correct atom is in the neighborhood of the selected multiple L indices. We propose a neighborhood-based MOLS (NMOLS) algorithm in this study. The NMOLS algorithm incorporates a neighborhood-based strategy to find the more matching atoms in each iteration. NMOLS selects the support from the candidate atoms by comparing their residual error. Simulation and image experimental results show that the proposed algorithm outperforms orthogonal matching pursuit (OMP), regularized OMP (ROMP), compressive sampling matching pursuit (CoSaMP), orthogonal least squares (OLS), and MOLS in terms of reconstruction accuracy.