Single-Frame Image Super-resolution Based on In-Place Patch Match Model

In this paper we propose a new high-quality and efficient single image super-resolution model that extends exploit the self-similarity property. The similarity of frequency error compensation between the high-resolution patch and low-resolution model can modeled as a optimization problem. Based on the in-place patch similarity, the optimization model is further simplified to alleviate the computing load. Experimental results shows that our algorithm can produce more natural structures in the sense that local similar pattern fusion can be regularized with the redundancy from theses in-place patches.