Web application for real-time enhancement of low-light images using Zero-DCE and hybrid optimization

This paper presents the development of a web application designed for the real-time enhancement of low-light images using advanced deep learning techniques, specifically Zero-DCE (Zero-Reference Deep Curve Estimation) and hybrid optimization methods. Low-light images often suffer from noise, poor visibility, and detail loss, making it challenging to interpret them effectively. Traditional image enhancement methods do not perform well in real-time applications. To address this, the proposed web application leverages the Zero-DCE model, which requires no reference image, making it suitable for diverse real-world conditions. Further, a hybrid optimization approach, combining techniques such as adaptive learning rate adjustments, GPU acceleration, and a Generative Adversarial Network (GAN) for fine-tuning, is employed to enhance computational efficiency, output quality, and perceptual realism. The application processes images in real-time, restoring detail, reducing noise, and improving contrast, thereby significantly enhancing the visibility of objects in low-light conditions. The system’s user-friendly interface allows for easy uploading and real-time processing of images, providing an accessible solution for various fields, including security, medical imaging, and satellite imagery, where clear visual interpretation is critical. We also evaluate the model’s performance on a diverse set of low-light images and compare it against state-of-the-art methods. An analysis of the computational requirements and user experience is also provided.