BSSRnet

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

Highlights:

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

2. Details of the Refinement Part.

3. Illustration of several kernels in bilateral filters.

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

Requirement

• PyTorch 1.3.0, torchvision 0.4.1. The code is tested with python=3.7, cuda=11.0.
• Matlab (For training/test data generation and performance evaluation)
• Prepare the train and test data following this.

Train

• Download the training sets from Baidu Drive (Key: NUDT) and unzip them to ./data/train/.
• Run train.py to perform training. Checkpoint will be saved to ./log/.

Test

• Download the test sets and unzip them to ./data/test/. Here, we provide the full test sets used in our paper on Google Drive and Baidu Drive (Key: NUDT).
• Run demo_test.py to perform a demo inference. Results (.png files) will be saved to ./results.

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}
}