Single image super-resolution using self-similarity and generalized nonlocal mean

摘要

In this paper, a super-resolution method based self-similarity and generalized nonlocal mean is proposed. The proposed method not only adopts the self-similarity of image to build a self-example training set but also exploits generalized nonlocal mean to improve the quality of the resultant image. In the proposed method, difference of Gaussians of the input low-resolution image is extracted firstly, and then a generalized nonlocal mean algorithm is proposed to estimate the missing high-frequency details of the low image. The experimental results show that the proposed algorithm has a good performance, and the high-resolution image generated by the proposed method is with better subjective and objective quality compared with other methods.