Home

Awesome

Variational-EM-based Deep Learning for Noise-blind Image Deblurring

This repo provides pre-trained models and the results on benchmark datasets of our CVPR 2020 paper. main paper, supp, poster

Usage

Download pretrained models. Put them into separate folders. The blurry inputs and kernels for Set12 can be found in this link.

Run test.py for deblurred images.

You can also test your data.

Network Structure

Results

Results on noise-blind deconvolution

Results on benchmark datasets

You can also download the deblurred results and run compute_metrics.py to compute the PSNR/SSIM with the same settings as ours. We also provide the results from FCNN as benchmark. Please also refer to their results.

Key References

IDD-BM3D: Danielyan, Aram, Vladimir Katkovnik, and Karen Egiazarian. "BM3D frames and variational image deblurring." IEEE Transactions on Image Processing 21.4 (2011): 1715-1728.

FDN: Kruse, Jakob, Carsten Rother, and Uwe Schmidt. " Learning to push the limits of efficient FFT-based image deconvolution. " Proceedings of the IEEE International Conference on Computer Vision. 2017.

EPLL-NA/GradNet7S: Jin, Meiguang, Stefan Roth, and Paolo Favaro. "Noise-blind image deblurring." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017.

DMSP: Bigdeli, Siavash Arjomand, et al. "Deep mean-shift priors for image restoration." Advances in Neural Information Processing Systems. 2017.

EPLL: Zoran, Daniel, and Yair Weiss. "From learning models of natural image patches to whole image restoration." 2011 International Conference on Computer Vision. IEEE, 2011.

CSF: Schmidt, Uwe, and Stefan Roth. "Shrinkage fields for effective image restoration." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2014.

FCNN: Zhang, Jiawei, et al. "Learning fully convolutional networks for iterative non-blind deconvolution." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017.

IRCNN: Zhang, Kai, et al. "Learning deep CNN denoiser prior for image restoration." Proceedings of the IEEE conference on computer vision and pattern recognition. 2017.

Bibtex

@InProceedings{Nan_2020_CVPR,
author = {Nan, Yuesong and Quan, Yuhui and Ji, Hui},
title = {Variational-EM-Based Deep Learning for Noise-Blind Image Deblurring},
booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2020}
}