Image reconstruction based on sparse and redundant representation model: Local vs nonlocal

This paper studied on image reconstruction techniques based on sparse and redundant representation in local and nonlocal ways. We expatiated on the principles of local and nonlocal reconstruction methods in sparse and redundant representation framework. Then we proposed a fixed point continuation solution for l1 regularization. We studied on the clustering-based sparse representation (CSR) algorithms, which combined dictionary learning and structure clustering in a unified variational framework. We used PSNR (Peak Signal to Noise Ratio) and MSSIM (Mean Structural Similarity) to evaluate the performance of those methods. Experimental results on different types of images indicate that combined local and nonlocal reconstruction model found the tradeoff between dictionary learning and structure clustering. It achieves the state of art reconstruction results and provides a valuable and promising reference for image reconstruction techniques.