Evaluating Low-Light Image Enhancement Across Multiple Intensity Levels

2025-11-19

AI Summary

Existing low-light image enhancement methods are constrained by paired training data captured under a single illumination condition, hindering rigorous evaluation of their generalization across diverse illumination levels. To address this, we introduce the Multi-Illumination Low-Light (MILL) dataset—the first benchmark explicitly designed for multi-illumination evaluation—comprising paired low-light/normal-light images acquired under fixed camera parameters with precise radiometric illumination calibration spanning a wide intensity range. Leveraging MILL, we systematically reveal substantial performance fluctuations of state-of-the-art methods across brightness levels—a previously uncharacterized limitation. To overcome the cross-illumination robustness bottleneck, we propose a structure-aware enhancement module. Experiments demonstrate that our method achieves up to 10 dB PSNR gain on DSLR-captured Full HD images and 2 dB on smartphone-captured ones, significantly improving adaptability and stability across heterogeneous lighting conditions.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Imaging in low-light environments is challenging due to reduced scene radiance, which leads to elevated sensor noise and reduced color saturation. Most learning-based low-light enhancement methods rely on paired training data captured under a single low-light condition and a well-lit reference. The lack of radiance diversity limits our understanding of how enhancement techniques perform across varying illumination intensities. We introduce the Multi-Illumination Low-Light (MILL) dataset, containing images captured at diverse light intensities under controlled conditions with fixed camera settings and precise illuminance measurements. MILL enables comprehensive evaluation of enhancement algorithms across variable lighting conditions. We benchmark several state-of-the-art methods and reveal significant performance variations across intensity levels. Leveraging the unique multi-illumination structure of our dataset, we propose improvements that enhance robustness across diverse illumination scenarios. Our modifications achieve up to 10 dB PSNR improvement for DSLR and 2 dB for the smartphone on Full HD images.

Problem

• Evaluating low-light image enhancement across multiple illumination intensity levels
• Addressing limited radiance diversity in existing low-light enhancement training datasets
• Improving algorithm robustness across variable lighting conditions through dataset innovation

Innovation

• Multi-illumination dataset enables cross-intensity evaluation
• Proposed modifications enhance robustness across illumination scenarios
• Achieved significant PSNR improvements for DSLR and smartphone