DeblurSDI: Blind Image Deblurring Using Self-diffusion

2025-10-31

AI Summary

Blind image deconvolution is a canonical ill-posed inverse problem requiring joint estimation of the latent sharp image and the unknown blur kernel. Existing approaches either rely on handcrafted priors or require large-scale external datasets for pretraining, limiting generalizability. This paper proposes a zero-shot self-supervised framework that operates without any pretraining: it formulates blind deconvolution as an iterative inverse self-diffusion process, achieving strong robustness to varying kernel sizes via a noise-scheduling mechanism while dynamically learning image-specific priors. A dual random-initialized network jointly optimizes the image and kernel, incorporating data consistency constraints and L1 sparsity regularization on the kernel. Experiments demonstrate stable recovery of high-fidelity images and accurate blur kernels across diverse severe degradation scenarios, consistently outperforming state-of-the-art methods.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Unsupervised & Self-Supervised Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Blind image deconvolution is a challenging ill-posed inverse problem, where both the latent sharp image and the blur kernel are unknown. Traditional methods often rely on handcrafted priors, while modern deep learning approaches typically require extensive pre-training on large external datasets, limiting their adaptability to real-world scenarios. In this work, we propose DeblurSDI, a zero-shot, self-supervised framework based on self-diffusion (SDI) that requires no prior training. DeblurSDI formulates blind deconvolution as an iterative reverse self-diffusion process that starts from pure noise and progressively refines the solution. At each step, two randomly-initialized neural networks are optimized continuously to refine the sharp image and the blur kernel. The optimization is guided by an objective function combining data consistency with a sparsity-promoting L1-norm for the kernel. A key innovation is our noise scheduling mechanism, which stabilizes the optimization and provides remarkable robustness to variations in blur kernel size. These allow DeblurSDI to dynamically learn an instance-specific prior tailored to the input image. Extensive experiments demonstrate that DeblurSDI consistently achieves superior performance, recovering sharp images and accurate kernels even in highly degraded scenarios.

Problem

• Solves blind image deblurring without external training data
• Estimates both sharp image and blur kernel simultaneously
• Uses self-diffusion process for robust kernel size adaptation

Innovation

• Zero-shot self-diffusion framework for blind deblurring
• Iterative reverse diffusion with dual-network optimization
• Dynamic noise scheduling for kernel-robust deconvolution