Unsupervised Ultra-High-Resolution UAV Low-Light Image Enhancement: A Benchmark, Metric and Framework

2025-09-01

AI Summary

Addressing severe image quality degradation, lack of paired training data, non-uniform illumination, and real-time deployment constraints in ultra-high-resolution (4K+) low-light UAV aerial imaging, this paper proposes U3LIE—the first end-to-end unsupervised solution for ultra-HD aerial low-light image enhancement. Methodologically, it introduces (1) U3D, the first unsupervised ultra-HD UAV low-light image dataset; (2) the Edge Efficiency Index (EEI), a novel metric jointly optimizing perceptual quality and hardware efficiency; (3) Adaptive Pre-Enhancement (APA) and Luminance Interval Loss (L_int) to enable effective unpaired training; and (4) a lightweight network architecture achieving real-time 4K image processing at 23.8 FPS on a single GPU. Quantitative and qualitative evaluations demonstrate state-of-the-art performance and strong deployability across diverse low-light aerial scenarios.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Machine Learning: Hardware-aware ML
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Low light conditions significantly degrade Unmanned Aerial Vehicles (UAVs) performance in critical applications. Existing Low-light Image Enhancement (LIE) methods struggle with the unique challenges of aerial imagery, including Ultra-High Resolution (UHR), lack of paired data, severe non-uniform illumination, and deployment constraints. To address these issues, we propose three key contributions. First, we present U3D, the first unsupervised UHR UAV dataset for LIE, with a unified evaluation toolkit. Second, we introduce the Edge Efficiency Index (EEI), a novel metric balancing perceptual quality with key deployment factors: speed, resolution, model complexity, and memory footprint. Third, we develop U3LIE, an efficient framework with two training-only designs-Adaptive Pre-enhancement Augmentation (APA) for input normalization and a Luminance Interval Loss (L_int) for exposure control. U3LIE achieves SOTA results, processing 4K images at 23.8 FPS on a single GPU, making it ideal for real-time on-board deployment. In summary, these contributions provide a holistic solution (dataset, metric, and method) for advancing robust 24/7 UAV vision. The code and datasets are available at https://github.com/lwCVer/U3D_Toolkit.

Problem

• Enhancing ultra-high-resolution UAV low-light images without supervision
• Addressing unique aerial challenges like non-uniform illumination and deployment constraints
• Providing a holistic solution with dataset, metric, and efficient framework

Innovation

• Unsupervised UHR UAV dataset with evaluation toolkit
• Edge Efficiency Index metric balancing quality and deployment
• Efficient framework with APA augmentation and Luminance Loss