Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising

13 Aug 2016Kai ZhangWangmeng ZuoYunjin ChenDeyu MengLei Zhang

Discriminative model learning for image denoising has been recently attracting considerable attentions due to its favorable denoising performance. In this paper, we take one step forward by investigating the construction of feed-forward denoising convolutional neural networks (DnCNNs) to embrace the progress in very deep architecture, learning algorithm, and regularization method into image denoising... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Image Super-Resolution BSD100 - 4x upscaling DnCNN-3 PSNR 27.29 # 21
Image Super-Resolution BSD100 - 4x upscaling DnCNN-3 SSIM 0.7253 # 26
Image Denoising BSD68 sigma10 DnCNN PSNR 36.31 # 2
Image Denoising BSD68 sigma15 DnCNN PSNR 31.73 # 3
Image Denoising BSD68 sigma25 DnCNN PSNR 29.23 # 3
Image Denoising BSD68 sigma30 DnCNN PSNR 30.40 # 2
Image Denoising BSD68 sigma50 DnCNN PSNR 26.23 # 6
Image Denoising BSD68 sigma70 DnCNN PSNR 26.56 # 2
Image Super-Resolution Set14 - 4x upscaling DnCNN-3 PSNR 28.04 # 24
Image Super-Resolution Set14 - 4x upscaling DnCNN-3 SSIM 0.7672 # 27
Image Super-Resolution Set5 - 4x upscaling DnCNN-3 PSNR 31.40 # 21
Image Super-Resolution Set5 - 4x upscaling DnCNN-3 SSIM 0.8845 # 24
Image Super-Resolution Urban100 - 4x upscaling DnCNN-3 PSNR 25.20 # 22
Image Super-Resolution Urban100 - 4x upscaling DnCNN-3 SSIM 0.7521 # 21
Image Denoising Urban100 sigma50 DnCNN PSNR 26.28 # 6
Image Denoising Urban100 sigma70 DnCNN PSNR 24.36 # 4