Image Denoising Based on Median Filter and UNet

Image denoising can improve image quality, improve visual perception, improve the accuracy of subsequent tasks, reduce image compression rate, and protect privacy information. In order to effectively remove the noise of real images, this paper proposes an image denoising method combining adaptive weighted median filtering algorithm, VGG and U-Net network. Firstly, the adaptive weighted median filtering algorithm is used to preprocess the input noise image, and then the VGG-16 network is used to extract the features of the image. On this basis, the U-Net network is constructed for encoding and decoding to obtain the denoised image. Experimental results on the SIDD real image dataset show that compared with other classical methods, the PSNR and SSIM values of the proposed algorithm are greatly improved, with PSNR reaching 35.57 and SSIM reaching 0.957. On the data sets with Gaussian noise standard deviation of 5, 10 and 20, the PSNR reaches 37.614, 35.241 and 35.249, SSIM reaches 0.974, 0.961 and 0.963, and MSE reaches 14.744, 31.985 and 36.120, respectively. The processed image has good visual effect, and the proposed algorithm can retain more image detail information while denoising.