Blind Universal Bayesian Image Denoising with Gaussian Noise Level Learning

5 Jul 2019  ·  Majed El Helou, Sabine Süsstrunk ·

Blind and universal image denoising consists of using a unique model that denoises images with any level of noise. It is especially practical as noise levels do not need to be known when the model is developed or at test time... We propose a theoretically-grounded blind and universal deep learning image denoiser for additive Gaussian noise removal. Our network is based on an optimal denoising solution, which we call fusion denoising. It is derived theoretically with a Gaussian image prior assumption. Synthetic experiments show our network's generalization strength to unseen additive noise levels. We also adapt the fusion denoising network architecture for image denoising on real images. Our approach improves real-world grayscale additive image denoising PSNR results for training noise levels and further on noise levels not seen during training. It also improves state-of-the-art color image denoising performance on every single noise level, by an average of 0.1dB, whether trained on or not. read more

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Grayscale Image Denoising BSD68 sigma10 BUIFD75 (blind) PSNR 33.47 # 2
Grayscale Image Denoising BSD68 sigma15 BUIFD75 (blind) PSNR 31.35 # 9
Grayscale Image Denoising BSD68 sigma20 BUIFD75 (blind) PSNR 29.88 # 1
Grayscale Image Denoising BSD68 sigma25 BUIFD75 (blind) PSNR 28.75 # 12
Grayscale Image Denoising BSD68 sigma30 BUIFD75 (blind) PSNR 27.82 # 2
Grayscale Image Denoising BSD68 sigma35 BUIFD75 (blind) PSNR 27.03 # 2
Grayscale Image Denoising BSD68 sigma40 BUIFD75 (blind) PSNR 26.31 # 1
Grayscale Image Denoising BSD68 sigma45 BUIFD75 (blind) PSNR 25.67 # 1
Grayscale Image Denoising BSD68 sigma5 BUIFD75 (blind) PSNR 37.25 # 1
Grayscale Image Denoising BSD68 sigma50 BUIFD75 (blind) PSNR 25.1 # 11
Grayscale Image Denoising BSD68 sigma55 BUIFD75 (blind) PSNR 24.55 # 1
Grayscale Image Denoising BSD68 sigma60 BUIFD75 (blind) PSNR 24.05 # 1
Grayscale Image Denoising BSD68 sigma65 BUIFD75 (blind) PSNR 23.56 # 1
Grayscale Image Denoising BSD68 sigma70 BUIFD75 (blind) PSNR 23.1 # 3
Grayscale Image Denoising BSD68 sigma75 BUIFD75 (blind) PSNR 22.67 # 2
Color Image Denoising CBSD68 sigma10 CBUIFD75 PSNR 35.98 # 1
Color Image Denoising CBSD68 sigma15 CBUIFD75 PSNR 33.66 # 4
Color Image Denoising CBSD68 sigma20 CBUIFD75 PSNR 32.02 # 1
Color Image Denoising CBSD68 sigma25 CBUIFD75 PSNR 30.76 # 4
Color Image Denoising CBSD68 sigma30 CBUIFD75 PSNR 29.71 # 1
Color Image Denoising CBSD68 sigma35 CBUIFD75 PSNR 28.81 # 3
Color Image Denoising CBSD68 sigma40 CBUIFD75 PSNR 28.01 # 1
Color Image Denoising CBSD68 sigma45 CBUIFD75 PSNR 27.28 # 1
Color Image Denoising CBSD68 sigma5 CBUIFD75 PSNR 40.05 # 1
Color Image Denoising CBSD68 sigma50 CBUIFD75 PSNR 26.60 # 9
Color Image Denoising CBSD68 sigma55 CBUIFD75 PSNR 25.95 # 1
Color Image Denoising CBSD68 sigma60 CBUIFD75 PSNR 25.34 # 1
Color Image Denoising CBSD68 sigma65 CBUIFD75 PSNR 24.75 # 1
Color Image Denoising CBSD68 sigma70 CBUIFD75 PSNR 24.18 # 1
Color Image Denoising CBSD68 sigma75 CBUIFD75 PSNR 23.63 # 3

Methods


No methods listed for this paper. Add relevant methods here