1 of 8

Blind Deconvolution

1

Traditional Blind Deconvolution [1990s-2015)

    • T. F. Chan and C.-K. Wong, “Total variation blind deconvolution,” TIP 1998
    • S. Cho and S. Lee, “Fast motion deblurring,” ACM ToG 2009
    • Cai, Jian-Feng, et al. "Blind motion deblurring from a single image using sparse approximation," CVPR 2009
    • L. Xu and J. Jia, “Two-phase kernel estimation for robust motion deblurring,” ECCV 2010

Image Estimation

Kernel Estimation

Conventional Alternating Minimization

Camera shake during exposure

+

=

Fronto-Parallel Scene

 

2 of 8

2

Conventional alternating minimization can be numerically unstable

Main Idea I – Something’s not right here

Image Estimation

Kernel Estimation

Iteration N

[1] Levin, Anat, et al. "Understanding blind deconvolution algorithms." IEEE TPAMI 2011

3 of 8

3

Main Idea II – A Toy Example

4 of 8

4

Main Idea III – So what should I do instead?

Fewer unknowns

5 of 8

5

Core Engine of our Methods

Dong, Jiangxin, et. al. "Deep wiener deconvolution: Wiener meets deep learning for image deblurring." NeurIPS 2020.

Non-Blind Solver: differentiable with respect to inputs

6 of 8

6

Blind Deconvolution with Diffusion Models

After each reverse diffusion step

Chung, Hyungjin, et al. "Parallel diffusion models of operator and image for blind inverse problems.“ CVPR 2023

7 of 8

7

Kernel Estimation Approach

8 of 8

8

Comparison with Blind-DPS

Metric

Blind-DPS

(Alternating Estimation)

Kernel-Diff (Ours)

(Kernel-First Estimation)

PSNR 🡩

17.56

19.07

SSIM 🡩

0.387

0.500

LPIPS 🡫

0.583

0.355

FID 🡫

280.53

172.33

Total Params

110M

43M

Time (in s)

337

305

Blurred image

Blind-DPS

Kernel-Diff