Neural Blind Deconvolution Using Deep Priors

[arxiv] [supp]

Introduction

Blind deconvolution is a classical yet challenging low-level vision problem with many real-world applications. Traditional maximum a posterior (MAP) based methods rely heavily on fixed and handcrafted priors that certainly are insufficient in characterizing clean images and blur kernels, and usually adopt specially designed alternating minimization to avoid trivial solution. In contrast, existing deep motion deblurring networks learn from massive training images the mapping to clean image or blur kernel, but are limited in handling various complex and large size blur kernels. Motivated by deep image prior (DIP) [1], we in this paper present two generative networks for respectively modeling the deep priors of clean image and blur kernel, and propose an unconstrained neural optimization solution to blind deconvolution (SelfDeblur). Experimental results show that our SelfDeblur can achieve notable quantitative gains as well as more visually plausible deblurring results in comparison to state-of-the-art blind deconvolution methods on benchmark datasets and real-world blurry images.

Prerequisites

• Python 3.6, PyTorch >= 1.0
• Requirements: opencv-python, tqdm
• Platforms: Ubuntu 16.04, cuda-10.0 & cuDNN v-7.5
• MATLAB for computing evaluation metrics

Datasets

SelfDeblur is evaluated on datasets of Levin et al. [2] and Lai et al. [3]. Please download the testing datasets from BaiduYun or OneDrive, and place the unzipped folders into ./datasets/.

Getting Started

1) Run SelfDeblur

Run shell scripts to deblur:

bash demo_levin.sh 
bash demo_lai.sh # coming
bash demo_real.sh # coming

All the deblurring results and deep models are also available. Please read results/levin/readme.docx and results/lai/readme.docx for the details. You can place the downloaded results into ./results/, and directly compute all the evaluation metrics in this paper.

2) Evaluation metrics

We also provide the MATLAB scripts to compute the average PSNR and SSIM values reported in the paper.

 cd ./statistic
 run statistic_levin.m 
 run statistic_lai.m 

SelfDeblur succeeds in simultaneously estimating blur kernel and generating clean image with finer texture details.

References

[1] D. Ulyanov, A. Vedaldi, and V. Lempitsky. Deep image prior. In IEEE CVPR 2018.

[2] A. Levin, Y. Weiss, F. Durand, and W. T. Freeman. Understanding and evaluating blind deconvolution algorithms. In IEEE CVPR 2009.

[3] W.-S. Lai, J.-B. Huang, Z. Hu, N. Ahuja, and M.-H. Yang. A comparative study for single image blind deblurring. In IEEE CVPR 2016.