Image Up-Scaling: Based on Self Similarity and Local Information

摘要

Most of the currently used image up-scaling algorithms are based on interpolation method. As the structural information is ignored, the result image will be blurred while sharp edges are interpolated smoothly. In our paper, we proposed a fast and local structure based algorithm, combined with improvement on up-sampling techniques, to enhance up-scaling image quality. The up-scaling process are guided through the self similarity computing. The structural information of image is retained effectively in this process. Experiments prove that our method has an obvious improvement on quality of result images, especially for details of texture rendering.