In order to solve the problem of missing high-frequency information in image super-resolution reconstruction by existing methods, this paper proposes a multi-layer nested residual network super-resolution reconstruction method based on attention mechanism, which uses different feature extraction schemes for different frequency information. The front-end feature information is directly transmitted to the back-end attention mechanism module by using the identity mapping connection across the residual network structure, and the hidden feature information in the original image is captured by adding the multi-layer nested residual network of the attention mechanism, and the image feature information is fused through the deep parallel residual network structure. Experimental results show that the improved algorithm can effectively improve the accuracy of image super-resolution reconstruction.