Physen-Noise2Noise: Physics-Guided Self-Supervised Defocus Deblurring with Bias Correction under Low-Light Conditions

2026-05-23

AI Summary

This work addresses the challenge of deblurring defocus-blurred images captured under low-light long-exposure conditions, which typically suffer from severe blur and complex biased noise. Conventional methods are limited by oversimplified noise models that fail to capture real-world imaging characteristics. To overcome this, the paper proposes a physics-guided self-supervised deblurring framework that, for the first time, incorporates frequency-domain constraints derived from defocus imaging physics into the Noise2Noise paradigm. The approach jointly models the physical image formation process and biased noise through learnable noise bias parameters, a multi-frame noisy initialization strategy, and a pretraining-finetuning mechanism. Notably, it achieves accurate noise bias correction and high-frequency detail recovery without requiring any clean reference images. Extensive experiments demonstrate that the proposed method significantly outperforms existing self-supervised approaches on both synthetic and real-world datasets.

Abstract

Low-light, long-exposure defocus deblurring remains a challenging problem due to the simultaneous presence of severe blur and complex biased noise. Existing methods typically rely on simplified noise assumptions, which limits their effectiveness under realistic imaging conditions. In this work, we propose Physen-Noise2Noise, a self-supervised deblurring framework guided by the physical model of defocus imaging, which leverages noisy multi-frame observations without requiring clean reference images. Unlike conventional Noise2Noise-based approaches that assume zero-mean noise, we derive a frequency-domain constraint inherent to the defocus imaging process and incorporate it into the learning framework via a learnable noise bias parameter. In addition, a multi-frame noisy initialization strategy is introduced to suppress complex biased noise prior to deblurring, providing a more stable starting point for reconstruction. This formulation explicitly models biased noise and enables joint bias correction and high-frequency detail recovery during training. Furthermore, we develop a pretrain-finetune variant to enhance robustness and generalization under challenging noise conditions. Extensive experiments on both simulation and real-world datasets demonstrate that the proposed method consistently outperforms state-of-the-art self-supervised approaches for defocus deblurring in the presence of complex biased noise.

Problem

• defocus deblurring
• low-light
• biased noise
• self-supervised learning
• noise modeling

Innovation

• physics-guided
• self-supervised
• defocus deblurring
• biased noise correction
• Noise2Noise