Noise Synthesis for Low-Light Image Denoising with Diffusion Models

2025-03-14

AI Summary

Low-light images suffer from non-Gaussian, non-stationary noise due to photon scarcity, and the scarcity of authentic paired data severely hinders denoising model training. To address this, we propose a diffusion-based synthesis method tailored for realistic low-light noise modeling. Our approach employs a dual-branch network to disentangle signal-dependent and signal-independent noise components, incorporates spatial position encoding to characterize sensor-specific fixed-pattern noise, and introduces an adaptive noise scheduling strategy that requires neither hardware calibration nor post-processing. This work presents the first end-to-end, interpretable modeling framework for low-light noise distributions. Synthesized noise exhibits high fidelity to real-world statistical properties across multiple benchmarks. Denoising networks trained on our synthetic data achieve state-of-the-art performance. Extensive noise decomposition and statistical validation further confirm the accuracy and physical plausibility of our noise model.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Low-light photography produces images with low signal-to-noise ratios due to limited photons. In such conditions, common approximations like the Gaussian noise model fall short, and many denoising techniques fail to remove noise effectively. Although deep-learning methods perform well, they require large datasets of paired images that are impractical to acquire. As a remedy, synthesizing realistic low-light noise has gained significant attention. In this paper, we investigate the ability of diffusion models to capture the complex distribution of low-light noise. We show that a naive application of conventional diffusion models is inadequate for this task and propose three key adaptations that enable high-precision noise generation without calibration or post-processing: a two-branch architecture to better model signal-dependent and signal-independent noise, the incorporation of positional information to capture fixed-pattern noise, and a tailored diffusion noise schedule. Consequently, our model enables the generation of large datasets for training low-light denoising networks, leading to state-of-the-art performance. Through comprehensive analysis, including statistical evaluation and noise decomposition, we provide deeper insights into the characteristics of the generated data.

Problem

• Addresses limitations of Gaussian noise models in low-light conditions.
• Proposes diffusion models for realistic low-light noise synthesis.
• Enables large dataset generation for training denoising networks.

Innovation

• Two-branch architecture for noise modeling
• Positional information for fixed-pattern noise
• Tailored diffusion noise schedule for precision