Image super-resolution refers to the reconstruction process of a high resolution image from one or a set of low resolution images.The theory of compressed sensing plays an important role in super-resolution reconstruction in recent years.However,existing super-resol ution models based on compressed sensing tend to have an initial estimation for the high-resolution i mage,while the quality of the estimation will directly affect the quality and efficiency of the recons truction.In this paper,the image nonlocal similarity is introduced as a regularization term into the neighb or embedding method, which improves the quality of the initial estimation.We incorporate the image m ulti-scale similarity and the local similarity into the reconstruction model and propose an image super-r esolution method based on neighbor embedding and structure self-similarity.The expression ability of p rior information is enhanced because of combining the advantages of the two aspects.The experiment uses the peak signal to noise ratio (PSNR) and the structural similarity index measure ment (SSIM) to compare the reconstructed image with the original image.Experimental results show that the proposed appro ach outperforms some state-of-the-art super-resolution methods in both subjective and objective e valuation criteria,which is able to enhance high frequency information in details and has a gain of 0.47－1. 84dB in PSNR on average.