Nighttime Traffic Surveillance Using Glare Reduction and Zero-DCE-Based Low-Light Image Enhancement

High-quality image plays a critical role in computer vision tasks such as object detection and scene understanding. However images captured in night traffic conditions usually suffer from low light and glare problems, which increases the difficulty of subsequent computer vision tasks. In this paper we propose a new approach to using Zero-DCE for image enhancement in low light on traffic surveillance cameras at night. We added a glare reduction (Reduce Glare) stage after the image enhancement process was carried out using Zero-DCE. This has been proven to help improve quality and maintain enhanced image structural information. The proof is the increase of the average PSNR from 28.91 to 31.07 and SSIM values from 0.55 to 0.88, for the enhanced images using the proposed approach compared to Zero-DCE alone. Subjective tests also carried out to see the condition of the image before and after processing. The outcomes of the suggested method were able to improve the image's brightness and contrast, visibility of details, and subjective visual quality. The impact can be felt directly by practitioners and parties involved in traffic monitoring, paving the way for more effective and reliable systems. These results not only overcome challenges in low-light conditions, but also open opportunities for the development of better surveillance technologies in the future.