Blind Image Deconvolution using Pretrained Generative Priors

TL;DR

This paper introduces a new blind image deconvolution method that uses pretrained deep generative models for sharp images and blur kernels, enabling effective deblurring even with significant noise and blur.

Contribution

It presents a novel alternating gradient descent approach in the latent space of pretrained generative models for blind image deblurring, incorporating classical priors for improved performance.

Findings

Excellent deblurring results under large blurs and noise

Effective in diverse image datasets with modifications

Combines deep generative and classical priors

Abstract

This paper proposes a novel approach to regularize the ill-posed blind image deconvolution (blind image deblurring) problem using deep generative networks. We employ two separate deep generative models - one trained to produce sharp images while the other trained to generate blur kernels from lower dimensional parameters. To deblur, we propose an alternating gradient descent scheme operating in the latent lower-dimensional space of each of the pretrained generative models. Our experiments show excellent deblurring results even under large blurs and heavy noise. To improve the performance on rich image datasets not well learned by the generative networks, we present a modification of the proposed scheme that governs the deblurring process under both generative and classical priors.