Home

Awesome

BSSRnet

PyTorch implementation of "Deep Bilateral Learning for Stereo Image Super-Resolution", IEEE Signal Processing Letters. <br><br>

Highlights:

1. We develop a bilateral dynamic network, which conduct space-variable filter on stereo images.

<p align="center"> <img src="https://github.com/xuqingyu26/BSSRnet/blob/main/Figs/Overview.png" width="100%"></p>

2. Details of the Refinement Part.

<p align="center"><img src="https://github.com/xuqingyu26/BSSRnet/blob/main/Figs/Refinement.png" width="100%"></p>

3. Illustration of several kernels in bilateral filters.

<p align="center"><img src="https://github.com/xuqingyu26/BSSRnet/blob/main/Figs/filter.png" width="100%"></p>

4. Our BSSR significantly outperforms PASSRnet with a comparable model size.

<p align="center"><img src="https://github.com/xuqingyu26/BSSRnet/blob/main/Figs/quantatitive.png" width="100%"></p>

Requirement

Train

Test

Citiation

We hope this work can facilitate the future research in stereo image SR. If you find this work helpful, please consider citing:

@article{xu2021deep,
  title={Deep Bilateral Learning for Stereo Image Super-Resolution},
  author={Xu, Qingyu and Wang, Longguang and Wang, Yingqian and Sheng, Weidong and Deng, Xinpu},
  journal={IEEE Signal Processing Letters},
  volume={28},
  pages={613--617},
  year={2021},
  publisher={IEEE}
}