Single Image Super-Resolution via Sparse Representation in Gradient Domain

Image super-resolution (SR) reconstruction is one of the most popular research topics in image processing for decades. This paper presents a novel approach to deal with single image SR problem. We search a mapping between a pair of low-resolution and high-resolution image patch in gradient domain by learning a generic image database and the input image itself. Given low-resolution image, the high-resolution image is reconstructed using sparse representation in gradient domain and solving Poisson equation. Experiments demonstrate that the state-of-the-art results have been achieved compared to other SR methods in terms of both PSNR and visual perception.