Local operator estimation for single-image super-resolution
Yi Tang, Hong Chen · 2015
The key of the problem of single-image super-resolution is the estimation of the relationship between low- and high-resolution images. In this paper, a novel single-image super-resolution algorithm is proposed which is motivated by the local manifold information of training samples and the structure information of image patches captured by matrix-value operators. By using the local manifold information of training samples, the similarities among low-resolution images are well estimated. Then, the structure information of image patches contained in the matrix-value operators provides the structure information of high-resolution image patches to the learning processes. By combining these information of image patches, the proposed single-image super-resolution algorithm achieves the state-of-the-art performance. Experimental results show the efficiency and the effectiveness of the proposed algorithm.