Blind Image Deconvolution by Generative-based Kernel Prior and Initializer via Latent Encoding

TL;DR

This paper introduces a novel blind image deconvolution framework that uses a generative adversarial network to model kernel priors and provide better initialization, significantly improving performance over existing DIP-based methods.

Contribution

It proposes a deep generative model-based approach for kernel prior modeling and initialization, enhancing the robustness and accuracy of blind image deconvolution.

Findings

Improved deblurring performance on multiple datasets.

High-quality kernel initialization via latent space encoding.

Enhanced robustness against non-convex optimization issues.

Abstract

Blind image deconvolution (BID) is a classic yet challenging problem in the field of image processing. Recent advances in deep image prior (DIP) have motivated a series of DIP-based approaches, demonstrating remarkable success in BID. However, due to the high non-convexity of the inherent optimization process, these methods are notorious for their sensitivity to the initialized kernel. To alleviate this issue and further improve their performance, we propose a new framework for BID that better considers the prior modeling and the initialization for blur kernels, leveraging a deep generative model. The proposed approach pre-trains a generative adversarial network-based kernel generator that aptly characterizes the kernel priors and a kernel initializer that facilitates a well-informed initialization for the blur kernel through latent space encoding. With the pre-trained kernel generator and initializer, one can obtain a high-quality initialization of the blur kernel, and enable optimization within a compact latent kernel manifold. Such a framework results in an evident performance improvement over existing DIP-based BID methods. Extensive experiments on different datasets demonstrate the effectiveness of the proposed method.