The paper aims to propose a super-resolution algorithm based on the self-similarity and sparse representation in combination with the self-similarity of images to solve the problem that the effect of the current sparse representation super-resolution algorithm depends on the training data. In the algorithm, the multi-dimensional self-similarity of images was used to construct amulti-dimensional image pyramid, and the improved similarity block search strategy was used to obtain the high and low resolution image blocks as training samples. The dictionary training was carried out to the samples. Finally, the super-resolution image was obtained according to sparse representation. The experimental results showed that the proposed algorithm was superior to other algorithms in peak signal to noise ratio (PSNR) and structural similarity (SSIM). For the experimental images, the average PSNR was increased by 0.5 dB. The proposed super-resolution algorithm does not need external database and has a good effect. It can be used for super-resolution reconstruction.