Multiscale Similarity Learning Single Image Super-resolution with Fast Edge Preserved Reconstruction

Super-resolution is an algorithm, which is capable of producing a high resolution output image with low resolution input image. This paper present a novel approach for producing high quality-high resolution image, which contain NE-based learning and fast edge filtered image reconstruction based on slope-limiter function. By using this approach, the produced high-resolution image maintain discontinuity across enhanced edges and preserve smoothly varying features. Experimental results show that the proposed method produce large PSNR value that of the state-of-the-art approaches.