Algorithm Design of Low-quality Image Restoration and Reconstruction for File Restoration
The traditional low-quality image restoration algorithm is difficult to learn the details of the image,which leads to the blurred edge of the repaired image,and there are some deficiencies in the detail information.To solve this problem,this paper analyzes the low-quality image,and proposes an image restoration and color reconstruction algorithm based on the improved deep convolutional generation adversarial network.The algorithm uses the coarse and fine scale network generator structure to replace the original model structure,in which the coarse-scale network can learn the global features,while the fine-scale net-work can learn the image edge details.The multi-scale attention mechanism is used to automatically fill the color,and then complete the image reconstruction.In the experimental test,the PSNR index,SSIM index and the running time of the pro-posed algorithm are better than the comparison algorithm,and the reconstructed image is rich in detail,so the algorithm has certain engineering practical value.
万方数据
维普
