Correction-Filter

Correction Filter for Single Image Super-Resolution: Robustifying Off-the-Shelf Deep Super-Resolvers (https://arxiv.org/abs/1912.00157 , Accepted to CVPR 2020 - oral)

Non-Blind

  1. Change the out/in directories in "correct_img.py".
  2. Define the sampling/reconstruction basis (s and r) in "correct_img.py".
  3. Set the scale factor in "Config.py".
  4. Run "correct_img.py"

Blind

  1. Define the SR network in "estimate_correction.py".
  2. Update the out/in directories in "estimate_correction.py"
  3. Set the scale factor in "Config.py".
  4. Run "estimate_correction.py"

Citation:

@ARTICLE{correction_filter,
  author = {{Abu Hussein}, Shady and {Tirer}, Tom and {Giryes}, Raja},
  title = "{Correction Filter for Single Image Super-Resolution: Robustifying Off-the-Shelf Deep Super-Resolvers}",
  journal = {In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  year = "2020"
}