Skip to content

Latest commit

 

History

History
47 lines (37 loc) · 2.15 KB

README.md

File metadata and controls

47 lines (37 loc) · 2.15 KB

Lightweight Image Super-Resolution with a Feature-Refined Network

This repository is Pytorch code for our proposed FRN. figure2 Schematic representation of the proposed Feature-Refined Network (FRN) and its submodules. The details about our proposed FRN can be found in our paper:https://www.sciencedirect.com/science/article/pii/S0923596522001771.

If you find our work useful in your research work, please star the code and consider citing:

@article{liu2023lightweight,
  title={Lightweight image super-resolution with a feature-refined network},
  author={Liu, Feiqiang and Yang, Xiaomin and De Baets, Bernard},
  journal={Signal Processing: Image Communication},
  volume={111},
  pages={116898},
  year={2023},
  publisher={Elsevier}
}

Requirements:

1. Python==3.6 (Anaconda is recommended)
2. skimage
3. imageio
4. Pytorch==1.2
5. tqdm
6. pandas
7. cv2 (pip install opencv-python)

Test:

python test.py -opt options/test/test_FRN_x2.json
python test.py -opt options/test/test_FRN_x3.json
python test.py -opt options/test/test_FRN_x4.json

Finally, PSNR/SSIM values for Set5 are shown on your screen, you can find the reconstruction images in ./results. Other standard SR benchmark dadasets, you need to
change the datasets storage path in the test_FRN_x2.json, test_FRN_x3.json and test_FRN_x4.json files.

Results:

Quantitative Results: Quantitative Results Average PSNR/SSIM for x2, x3 and x4 SR on datasets Set5, Set14, B100, Urban100, and Manga109. The best and second best results are highlighted in red and blue, respectively.

Some Qualitative Results: figure5 Visual comparison of the results of our FRN with those of other state-of-the-art lightweight methods on some images from the B100 and Urban100 datasets for x4 SR. The best results are indicated in bold.