Super-resolution via a patch-based sparse algorithm

Abstract

The Sparsity concept has been widely used in image processing applications. In this paper, an approach for super-resolution has been proposed which uses sparse transform. This approach has mixed the inpainting concept with zooming via a sparse representation. A dictionary is being trained from a low-resolution image and then a zoomed version of this low resolution image will use that dictionary in a few iterations to fill the undefined image pixels. Experimental results confirm the strength of this algorithm against the other interpolation algorithms.