Official Code for Structured Kernel Estimation for Photon-Limited Deconvolution (CVPR 2023)

Instructions

1. Create a local copy of repository using the following commands

   foor@bar:~$ git clone https://github.com/sanghviyashiitb/structured-kernel-cvpr23.git
   foor@bar:~$ cd structured-kernel-cvpr23
   foor@bar:~/structured-kernel-cvpr23$       

2. Download the pretrained models, i.e. denoiser, p4ip, and ktn into model_zoo from the link here

3. To test the network on levin-data, run the file

   foor@bar:~/structured-kernel-cvpr23$ python3 demo_grayscale.py  

   

4. To test the network on real-sensor data, run the file

   foor@bar:~/structured-kernel-cvpr23$ python3 demo_real.py  

   

For further details on this dataset containing real-sensor noise + motion blur along with ground-truth kernels i.e., Photon Limited Deblurring Dataset (PLDD) refer to this link

Citation

@InProceedings{Sanghvi_2023_CVPR,
   author    = {Sanghvi, Yash and Mao, Zhiyuan and Chan, Stanley H.},
   title     = {Structured Kernel Estimation for Photon-Limited Deconvolution},
   booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
   month     = {June},
   year      = {2023},
   pages     = {9863-9872}
}

Feel free to ask your questions/share your feedback using the Issues feature at the the top of this page.