Image compressive sensing reconstruction based on cooperative adversarial optimization network

In view of the information loss in existing deep network-based image compressive sensing reconstruction algorithm, which leads to blurred reconstruction, a generative adversarial mechanism is introduced, and an image compressive sensing reconstruction algorithm based on cooperative adversarial optimization network is proposed. The algorithm inputs the image observations into the generator, and uses the multi-scale structure feature extraction module to refine the multi-level structure of the image; introduces an adversarial mechanism, and uses the non-local similar features of the image to confront the generated images of the generator. , to achieve accurate reconstruction of the original image. The experimental results show that compared with the existing algorithm, the objective evaluation index of the reconstructed image increases the PSNR value by 1.68 - 2.33 dB, and the SSIM by 0.037 6 - 0.059 2. The visual effect of the image is outstanding, and it can effectively construct finer image features.