Reviving Darkness: Enhancing Low Visibility Images in Adverse Conditions using an Optimized Zero-DCE Model

Capturing clear images in low-light environments such as fog, rain, and moonlight is a major challenge, impacting applications like surveillance, autonomous driving, and nighttime navigation. Existing deep learning-based methods, including Zero-Reference Deep Curve Estimation (Zero-DCE), often struggle with over-saturation, unnatural color balance, noise amplification, and loss of structural details. This research introduces an Optimized Zero-DCE Model that specifically addresses these limitations by refining the loss function to achieve a balance between brightness enhancement, noise suppression, and color consistency. The model incorporates adaptive parameter tuning and an improved training strategy to prevent over-enhancement and retain fine details. Experimental results show that the optimized model significantly enhances image quality, achieving a Peak Signal-to-Noise Ratio (PSNR) of 21.90 and a Structural Similarity Index (SSIM) of 0.9469, outperforming conventional enhancement techniques. These improvements enable more reliable low-light vision for safety-critical applications, ensuring better clarity and usability in extreme environmental conditions.