STPAN: Learning spatial and spatio-temporal pixel aggregations for image and video denoising

This is a official implementation of the paper Learning Spatial and Spatio-Temporal Pixel Aggregations for Image and Video Denoising, TIP 2020.

An earlier arXiv paper: Learning Deformable Kernels for Image and Video Denoising, 2019.

The pytorch implementation is still under preparation and will be released soon.

Please see tf branch for original implementation and models.

Thanks to z-bingo for re-implementing our method in https://github.com/z-bingo/Deformable-Kernels-For-Video-Denoising.git

TODOs

• Release SPAN and PAN code
• Release Test and Training Scripts
• Release Datasets
• Release Video Denoise models
• Release Image Denoise models

Installation

For environment setting up:

pip install -r requirement.txt

Dataset Preparation

Training

Testing

Citation

If you find this repo helpful in your research, please cite our paper.

@article{stpan,
  title={Learning spatial and spatio-temporal pixel aggregations for image and video denoising},
  author={Xu, Xiangyu and Li, Muchen and Sun, Wenxiu and Yang, Ming-Hsuan},
  journal={IEEE Transactions on Image Processing},
  year={2020}
}
@article{dcn_denoise,
  title={Learning deformable kernels for image and video denoising},
  author={Xu, Xiangyu and Li, Muchen and Sun, Wenxiu},
  journal={arXiv preprint arXiv:1904.06903},
  year={2019}
}

Contact: muchenli1997@gmail.com