LL-Gaussian: Low-Light Scene Reconstruction and Enhancement via Gaussian Splatting for Novel View Synthesis

2025-04-14

AI Summary

Existing NeRF-based methods for novel view synthesis (NVS) of low-light sRGB images suffer from high computational cost and sensitivity to input quality, while 3D Gaussian Splatting (3DGS) exhibits unstable initialization and unreliable enhancement under severe noise and low dynamic range. Method: We propose the first end-to-end 3DGS reconstruction and enhancement framework tailored for extremely low-light scenes. It features a low-illumination-adaptive Gaussian initialization module, a reflection–illumination-decoupled dual-branch Gaussian representation model, and an unsupervised optimization strategy jointly driven by physical constraints and diffusion priors. Contribution/Results: Evaluated on a newly constructed extremely low-light dataset, our method significantly improves geometric accuracy and rendering quality. It reduces training time to just 2% of state-of-the-art NeRF methods and achieves 2000× faster inference, enabling high-fidelity real-time NVS under pseudo-normal illumination conditions.

Technology Category

• Computer Vision: 3D Computer Vision
• Search and Optimization: Learning to Search
• Machine Learning: Learning with Manifolds

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Novel view synthesis (NVS) in low-light scenes remains a significant challenge due to degraded inputs characterized by severe noise, low dynamic range (LDR) and unreliable initialization. While recent NeRF-based approaches have shown promising results, most suffer from high computational costs, and some rely on carefully captured or pre-processed data--such as RAW sensor inputs or multi-exposure sequences--which severely limits their practicality. In contrast, 3D Gaussian Splatting (3DGS) enables real-time rendering with competitive visual fidelity; however, existing 3DGS-based methods struggle with low-light sRGB inputs, resulting in unstable Gaussian initialization and ineffective noise suppression. To address these challenges, we propose LL-Gaussian, a novel framework for 3D reconstruction and enhancement from low-light sRGB images, enabling pseudo normal-light novel view synthesis. Our method introduces three key innovations: 1) an end-to-end Low-Light Gaussian Initialization Module (LLGIM) that leverages dense priors from learning-based MVS approach to generate high-quality initial point clouds; 2) a dual-branch Gaussian decomposition model that disentangles intrinsic scene properties (reflectance and illumination) from transient interference, enabling stable and interpretable optimization; 3) an unsupervised optimization strategy guided by both physical constrains and diffusion prior to jointly steer decomposition and enhancement. Additionally, we contribute a challenging dataset collected in extreme low-light environments and demonstrate the effectiveness of LL-Gaussian. Compared to state-of-the-art NeRF-based methods, LL-Gaussian achieves up to 2,000 times faster inference and reduces training time to just 2%, while delivering superior reconstruction and rendering quality.

Problem

• Reconstructing and enhancing low-light scenes for novel view synthesis
• Overcoming unstable Gaussian initialization from low-light sRGB inputs
• Reducing computational costs and training time compared to NeRF methods

Innovation

• End-to-end Low-Light Gaussian Initialization Module (LLGIM)
• Dual-branch Gaussian decomposition model
• Unsupervised optimization with physical and diffusion guidance