Image Super-Resolution Reconstruction via EROMP Sparse Representation

Abstract

In order to improve the resolution of single-image, a new super-resolution reconstruction method is proposed using sparse representation via enhanced regularized-orthogonal-matching-pursuit. The core task of the SR problem is to solve the basis representation of image patches with respect to corresponding over-complete dictionary. Since the guarantee and the speed of a coding algorithm are very important in both dictionary learning and signal decomposition, we present a rapid sparse representation algorithm. Moreover, only low resolution dictionary is learned from image examples for reducing time consumption of dictionary learning. And the correspondence of high resolution is obtained under the numerical calculation. Experimental results show that the proposed method can effectively improve image resolution. The peak signal to noise ratio and structural similarity are gained 2.1dB and 0.09 respectively, compared with Bicubic interpolation widely used.