Super-resolution Reconstruction Algorithm Based on Patch Similarity and Back-projection Modification

摘要

We propose an effective super resolution reconstruction algorithm based on patch similarity and back-projection modification. In the proposed algorithm, we assume patch to be similar in natural images and extract the high-frequency information from the best similar patch to add into goal high-resolution image. In the process of reconstruction, the high-resolution patch is back-projected into the low-resolution patch so as to gain detailed modification. Experiments performed on simulated low-resolution image and real low-resolution image are proved that the proposed super-resolution reconstruction algorithm is effective and efficient to improve the resolution of image and achieve a better visual performance.