FormResNet-Denoise-Gaussian-noise-TensorFlow
Simplely implement the paper 'FormResNet: Formatted Residual Learning for Image Restoration' by TensorFlow
Introduction
This code just simplely implement the paper FormResNet: Formatted Residual Learning for Image Restoration, but there are some details of the code are different from the paper.
DataSets
The datasets include 400 gray images, but i have croped them into 40x40 patches. the croped datasets can be downloaded from my BaiduYun
Examples of TrainingSet
 |
 |
 |
 |
 |
 |
 |
 |
Python packages
====================
- python3.5
- tensorflow1.4.0
- pillow
- numpy
- scipy
- skimage
====================
Results of the code
Trained about 1 epoch, noise intensity sigma: 25
| Raw | Noised | FormResNet[1] | DnCNN[2] |
|---|---|---|---|
 |
 |
 |
 |
| - | - | psnr/ssim 26.12/0.82 | psnr/ssim 25.01/0.77 |
 |
 |
 |
 |
| - | - | psnr/ssim 32.01/0.83 | psnr/ssim 31.24/0.80 |
 |
 |
 |
 |
| - | - | psnr/ssim 25.55/0.86 | psnr/ssim 24.53/0.82 |
 |
 |
 |
 |
| - | - | psnr/ssim 27.66/0.85 | psnr/ssim 26.30/0.82 |
 |
 |
 |
 |
| - | - | psnr/ssim 29.392/0.90 | psnr/ssim 28.85/0.87 |
 |
 |
 |
 |
| - | - | psnr/ssim 28.34/0.84 | psnr/ssim 28.00/0.80 |
 |
 |
 |
 |
| - | - | psnr/ssim 22.00/0.83 | psnr/ssim 23.83/0.79 |
[1] Jiao J, Tu W C, He S, et al. FormResNet: Formatted Residual Learning for Image Restoration[C]// Computer Vision and Pattern Recognition Workshops. IEEE, 2017:1034-1042.
[2] Zhang K, Zuo W, Chen Y, et al. Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising[J]. IEEE Transactions on Image Processing A Publication of the IEEE Signal Processing Society, 2017, 26(7):3142-3155.