Multi-variable intelligent matching pursuit algorithm using prior knowledge for image reconstruction by l0 minimization

摘要: Image reconstruction by l0 minimization is an NP-hard problem that requires exhaustively listing all possibilities of the original signal with a very high computational complexity, which difficult to be achieved traditional algorithms. Although greedy algorithm aims at solving minimization, it more likely fall into suboptimal solution. In this paper, we propose multi-variable intelligent matching pursuit (MIMP), can solve essentially taking advantage optimization in combinatorial problems and searching for global optimal solution improve performance image reconstruction. The updating mechanism MIMP designed introducing strategies accelerate speed. Also, scheme utilized sample images then joint implemented measurements, not only accuracy but also reduce complexity. Moreover, edge saliency obtained as prior knowledge guide compressive sensing reconstruction, contributes lot complexity As sparsity level hard estimated, new function proposed without knowing prior. Compared other state-of-the-art algorithms, method achieve better has reasonable relatively faster speed using knowledge. Numerical experiments on several demonstrate significantly outperforms algorithms structure based PSNR, SSIM visual quality. HighlightsWe significantly.A novel unknown prior.Multi-variable introduced small measurement number.Edge information applied computation performance.