README.md

GLSGN: Global-Local Stepwise Generative Network

Official pytorch codes for the paper:

• Global-Local Stepwise Generative Network for Ultra High-Resolution Image Restoration

The code and dataset will be released here once our paper is accepcted.

We conduct experiments on three challenging image restoration tasks: image reflection removal, image deraining, and image dehazing.

Network Architecture

The overall framework of our proposed GLSGN.

Laplacian Pyramid

UHR4K Dataset

This dataset includes three parts: UHR4K-Syn, UHR4K-Real, and UHR4k-Rain.

Citation

If you find this work useful for your research, please cite:

@article{feng2022global,
  title={Global-Local Stepwise Generative Network for Ultra High-Resolution Image Restoration},
  author={Feng, Xin and Ji, Haobo and Pei, Wenjie, and Lu, Guangming},
  journal={arXiv preprint arXiv:2207.08808},
  year={2022}
}