Image Super-resolution Reconstruction Based on Multi-layer Parallel Residual Network

The existing methods can improve the overall visual effect of the image, but the reconstructed image is still lack of high-frequency information, leading to texture blur. To solve these problems, this paper proposes an image super-resolution reconstruction algorithm based on multi-layer parallel convolution and residual network. Taking the multilayer parallel structure as the whole framework, firstly, different convolution combinations are used to enrich the feature information of the parallel structure, and the jump connection is added to further enrich the feature and fuse the output to extract more high-frequency information. Secondly, an adaptive residual network is introduced to supplement information and optimize network performance. Finally, perceptual loss is used to improve the overall quality of the restored image. Experimental results show that the proposed algorithm performs well for the reconstructed image, and the PSNR and structural similarity are significantly improved in the objective evaluation.