/MN-A-model-driven-network-for-guided-image-denoising

A model-driven network for guided image denoising

Primary LanguagePython

A model-driven network for guided image denoising

Shuang Xu, Jiangshe Zhang, Jialin Wang, Kai Sun*, Chunxia Zhang, Junmin Liu, Junying Hu

Test

  1. Download all the files from this hyperlink.
  2. Double-click dataset.zip to extract data.
  3. Download the network weights from Pretrained weights and place .pth files in ./weight folder.
  4. Run test_MN.py to reproduce the following tables.
(a) RNS (b) FAIP

The table shows the evaluation results on RNS and FAIP with Gaussian noise. The 1st, 2nd and 3rd best values are marked by bold, red and underline, respectively.

Test on your own images

  1. Place your images in ./eval/Flash/guidance and ./eval/Flash/target.
  2. Download the network weights from Pretrained weights and place .pth files in ./weight folder.
  3. Run eval_MN.py to reproduce the following images.
Guidance Target Denoised

Here are tested results on real-world .

Pretrained weights

Model # layers # filters Modality
MN 7 64 RGB-NIR
MN 7 64 Nonflash-Flash
MN-L 3 32 RGB-NIR
MN-L 3 32 Nonflash-Flash

Citation

@article{xu2022_MN,
     author = {S. Xu, J. Zhang, J. Wang, K. Sun, C. Zhang, J. Liu, J. Hu},
     title = {A model-driven network for guided image denoising},
     journal = {Inf. Fus.},
     volume = {85},
     number = {},
     pages = {60--71},
     month = {Sep.},
     year = {2022},
}