Image super-resolution with facet improvement and detail enhancement based on local self examples

Abstract

In this paper, an algorithm named ILSE (Improved Local Self Examples) is proposed to improve the facet phenomenon based on [1]. The algorithm divides the high frequency component into two orthogonal spaces by analyzing the covariance matrix of a single image. Moreover, the high resolution image can be further improved by using the IWLS (Improved Weighted Least Square) filter, in which we add one more regulation term to balance the gradient field and 2-order holomorphic complete differential form of the image. The algorithm can preserve the edge sharpness and enhance the details at the same time. Analysis is given on the relations and the differences between the proposed approach and some other state-of-the-art interpolation methods. Experimental results show that the proposed method can achieve better image quality as compared to other competitors.