Sparse Representation Based Image Super-resolution Combining Rotation Strategy and Nonlocal Self-similarity
Di Wang, Juan Li, Jianliang Chen · 2021
This paper proposes an improved image super-resolution method based on sparse representation. A pair of dictionaries are learned by minimizing the sparse recovery error of high-resolution sample patches. In the reconstruction phase, the learned dictionaries and residual compensation are used to reconstruct the high-resolution image. In order to further reduce the sparse recovery error, a rotation strategy is introduced. Moreover, the image nonlocal self-similarity is exploited to constrain super-resolution reconstruction. Experimental results show that the proposed method can effectively improve PSNR, SSIM and visual effects.