LoLI-Street: Benchmarking Low-Light Image Enhancement and Beyond

Md Tanvir Islam,Inzamamul Alam,Simon S. Woo,Saeed Anwar,IK Hyun Lee,Khan Muhammad

2024-10-13

Abstract:Low-light image enhancement (LLIE) is essential for numerous computer vision tasks, including object detection, tracking, segmentation, and scene understanding. Despite substantial research on improving low-quality images captured in underexposed conditions, clear vision remains critical for autonomous vehicles, which often struggle with low-light scenarios, signifying the need for continuous research. However, paired datasets for LLIE are scarce, particularly for street scenes, limiting the development of robust LLIE methods. Despite using advanced transformers and/or diffusion-based models, current LLIE methods struggle in real-world low-light conditions and lack training on street-scene datasets, limiting their effectiveness for autonomous vehicles. To bridge these gaps, we introduce a new dataset LoLI-Street (Low-Light Images of Streets) with 33k paired low-light and well-exposed images from street scenes in developed cities, covering 19k object classes for object detection. LoLI-Street dataset also features 1,000 real low-light test images for testing LLIE models under real-life conditions. Furthermore, we propose a transformer and diffusion-based LLIE model named "TriFuse". Leveraging the LoLI-Street dataset, we train and evaluate our TriFuse and SOTA models to benchmark on our dataset. Comparing various models, our dataset's generalization feasibility is evident in testing across different mainstream datasets by significantly enhancing images and object detection for practical applications in autonomous driving and surveillance systems. The complete code and dataset is available on <a class="link-external link-https" href="https://github.com/tanvirnwu/TriFuse" rel="external noopener nofollow">this https URL</a>.

Computer Vision and Pattern Recognition,Artificial Intelligence,Computational Engineering, Finance, and Science

What problem does this paper attempt to address?

This paper attempts to solve several key problems in the field of Low - Light Image Enhancement (LLIE):

1. **Lack of paired datasets for street scenes**:
- The existing LLIE datasets are limited in quantity and diversity, especially those for urban street scenes are very scarce. This restricts the performance of LLIE models in practical applications, especially in autonomous vehicles and surveillance systems.

2. **Poor performance of existing methods in real low - light conditions**:
- Although some advanced transformer and diffusion models perform well on synthetic datasets, when dealing with low - light images in the real world, the effects of these methods are still not satisfactory. They are difficult to maintain high precision and robustness in complex real - life scenarios.

3. **Improving the practical application effects of low - light image enhancement**:
- In order to improve the performance of computer vision systems in low - light environments, especially for application scenarios such as autonomous vehicles and surveillance systems, more effective LLIE methods need to be developed. These problems not only affect the convenience and efficiency of the system, but also are directly related to safety and reliability.

To solve the above problems, the author proposes and contributes the following:

- **Introducing a new dataset "LoLI - Street"**:
- It contains 33,000 paired low - light and normal - light images, covering the detection tasks of 19,000 object categories. In addition, it also includes 1,000 real low - light test images for evaluating the performance of LLIE models under actual conditions.

- **Proposing a new LLIE model "TriFuse"**:
- It combines the advantages of transformer and diffusion models, improves efficiency by reducing the sampling steps in the diffusion process, and enhances image quality and detail preservation through the Conditional Noise Module (CNM) and Edge Sharpening Module (ESM).

Through these contributions, the author aims to promote the research progress in the field of LLIE, especially to achieve better low - light image enhancement effects in practical applications.