Evaluation and Investigation on the Blur Medical Image Deblurring Algorithm Based on Sports

Clear medical images can greatly assist doctors in pathological analysis and disease diagnosis. However, due to factors such as instrument equipment and environmental conditions, the collected medical images may sometimes exhibit some blurry phenomena, which in turn can lead to difficulties in diagnostic analysis. Therefore, image deblurring technology has emerged and is widely used in fields such as traffic monitoring, medical imaging, and photography. Although there are many methods for image deblurring at present, most of them have drawbacks such as high computational complexity, noise in restored images, and ringing effects. Therefore, based on the generative adversarial networks (GANs), this article improved the ResNet module to address the problems of high training difficulty and complex parameter debugging and obtained a simpler deep learning network model, that is, Deblur ResNet. Finally, SSIM and PSNR values were used as evaluation criteria, and a comparative experiment was conducted between the algorithm proposed in this paper and the two deblurring algorithms, DeblurGAN and RDN, which had good results. The experimental results showed that compared with DeblurGAN, the algorithm proposed in this paper improved PSNR and SSIM by 14.9% and 7.9%, respectively; compared with RDN, it increased by 10.1% and 5.2% in PSNR and SSIM indicators. This result indicated that the proposed model achieved good deblurring performance on the dataset.