LENVIZ: A High-Resolution Low-Exposure Night Vision Benchmark Dataset

2025-03-25

AI Summary

Low-light image enhancement is critical for night vision, surveillance, and autonomous driving, yet hindered by the lack of high-quality, large-scale benchmarks. To address this, we introduce LowLight-4K—the first large-scale, multi-exposure, 4K-resolution low-light night-vision benchmark—comprising 230,000 real-world indoor/outdoor frames captured synchronously via a triple-sensor setup, covering diverse illumination conditions, noise patterns, and semantic complexity. Ground-truth images are meticulously retouched by professional photographers to ensure photorealistic fidelity. LowLight-4K uniquely combines large-scale multi-exposure acquisition with human-refined ground truth and provides a comprehensive, standardized evaluation framework for state-of-the-art methods. Extensive experiments reveal systematic limitations across prevailing approaches in dynamic range recovery, texture preservation, and the noise-structure trade-off. As the largest publicly available 4K low-light enhancement benchmark, LowLight-4K establishes a robust foundation for algorithm development and rigorous performance assessment.

Technology Category

• Computer Vision: Low Level & Physics-based Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Low-light image enhancement is crucial for a myriad of applications, from night vision and surveillance, to autonomous driving. However, due to the inherent limitations that come in hand with capturing images in low-illumination environments, the task of enhancing such scenes still presents a formidable challenge. To advance research in this field, we introduce our Low Exposure Night Vision (LENVIZ) Dataset, a comprehensive multi-exposure benchmark dataset for low-light image enhancement comprising of over 230K frames showcasing 24K real-world indoor and outdoor, with-and without human, scenes. Captured using 3 different camera sensors, LENVIZ offers a wide range of lighting conditions, noise levels, and scene complexities, making it the largest publicly available up-to 4K resolution benchmark in the field. LENVIZ includes high quality human-generated ground truth, for which each multi-exposure low-light scene has been meticulously curated and edited by expert photographers to ensure optimal image quality. Furthermore, we also conduct a comprehensive analysis of current state-of-the-art low-light image enhancement techniques on our dataset and highlight potential areas of improvement.

Problem

• Enhancing low-light images for night vision and surveillance
• Addressing challenges in low-illumination image capture
• Providing a high-resolution benchmark dataset for improvement

Innovation

• Introduces LENVIZ high-resolution night vision dataset
• Includes 230K frames with expert-curated ground truth
• Evaluates state-of-the-art low-light enhancement techniques