Gaussian Light Field Splatting: A Physical Prior-Driven Vision Transformer for Unsupervised Low-Light Image Enhancement

2026-06-16

AI Summary

This work addresses the challenges of uneven local exposure and color distortion in unsupervised low-light image enhancement by proposing a Vision Transformer framework integrated with physical priors. It introduces Gaussian light field splatting into the Transformer architecture for the first time, modeling non-uniform illumination through anisotropic Gaussian kernels and embedding physics-guided bias into the self-attention mechanism to adaptively infer spatial gain fields. To further enhance fidelity, the method incorporates a chrominance vector angular loss and a luminance edge-aware loss, which jointly preserve color accuracy and structural details. Experimental results demonstrate that the proposed approach significantly outperforms existing methods in both illumination correction and detail retention, achieving state-of-the-art performance in unsupervised low-light image enhancement.

Abstract

Existing unsupervised low-light image enhancement methods often encounter local exposure imbalance and color distortion under complex non-uniform illumination. In addition, most Vision Transformers lack an explicit mechanism for modeling the physical priors of illumination degradation. To address these limitations, we propose GLFS, a Gaussian light field splatting-based Vision Transformer that integrates continuous physical illumination modeling from Gaussian splatting into the Transformer architecture. In GLFS, scene illumination is represented by a superposition of anisotropic Gaussian basis functions. Physics-guided biases are introduced into self-attention to adaptively infer a spatial gain field, enabling accurate and uniform restoration under complex illumination. To reduce color bias and structural degradation during enhancement, a color-vector angular loss and a luminance-edge loss are further developed. These losses enforce hue consistency and improve the structural fidelity of local details. Extensive ablation studies and quantitative evaluations show that GLFS provides clear advantages in illumination correction and detail preservation. It achieves state-of-the-art performance and offers a new representation paradigm for low-light image enhancement.

Problem

• low-light image enhancement
• exposure imbalance
• color distortion
• non-uniform illumination
• physical priors

Innovation

• Gaussian splatting
• Vision Transformer
• physical prior
• unsupervised low-light enhancement
• spatial gain field