A Poisson-Guided Decomposition Network for Extreme Low-Light Image Enhancement

2025-06-04

AI Summary

To address the challenge of enhancing images corrupted by signal-dependent Poisson noise under extremely low-light conditions, this paper proposes a lightweight end-to-end Retinex decomposition network. Methodologically, it is the first to embed Poisson noise modeling into the Retinex framework, enabling prior-free illumination estimation and reflectance recovery via joint learning of illumination and reflectance; a Poisson-aware loss function is further designed to jointly optimize denoising and brightness enhancement. The key contributions are: (1) eliminating the common Gaussian noise assumption and handcrafted priors, thereby preserving color constancy and structural fidelity; (2) achieving integrated enhancement with smooth illumination maps, chromaticity bias suppression, and fine-detail preservation; (3) achieving state-of-the-art PSNR/SSIM performance on multiple extreme low-light benchmarks while supporting real-time inference.

Technology Category

Computer Vision: Computational Photography, Image & Video SynthesisSearch and Optimization: Learning to SearchMachine Learning: Calibration & Uncertainty Quantification

Application Category

Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for rankingSecurity and Privacy: Privacy-enhancing technologiesUser Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Low-light image denoising and enhancement are challenging, especially when traditional noise assumptions, such as Gaussian noise, do not hold in majority. In many real-world scenarios, such as low-light imaging, noise is signal-dependent and is better represented as Poisson noise. In this work, we address the problem of denoising images degraded by Poisson noise under extreme low-light conditions. We introduce a light-weight deep learning-based method that integrates Retinex based decomposition with Poisson denoising into a unified encoder-decoder network. The model simultaneously enhances illumination and suppresses noise by incorporating a Poisson denoising loss to address signal-dependent noise. Without prior requirement for reflectance and illumination, the network learns an effective decomposition process while ensuring consistent reflectance and smooth illumination without causing any form of color distortion. The experimental results demonstrate the effectiveness and practicality of the proposed low-light illumination enhancement method. Our method significantly improves visibility and brightness in low-light conditions, while preserving image structure and color constancy under ambient illumination.

Problem

• Denoising images degraded by Poisson noise in extreme low-light conditions
• Enhancing illumination while suppressing signal-dependent noise effectively
• Preserving image structure and color constancy without distortion

Innovation

• Poisson denoising loss for signal-dependent noise
• Retinex-based decomposition in encoder-decoder network
• Light-weight deep learning for illumination enhancement