Performance Study of Enhancing Low Light Images using Deep Vision Networks

Improving image quality in low-light conditions is difficult and frequently leads to grainy and noisy photos. Traditional picture restoration approaches suffer due to a lack of visual data. Recent advances in deep learning show potential for addressing this problem. The goal of this research is to create and compare three cutting-edge image restoration techniques that are specifically designed for low-light circumstances. One of these methods employs the MIRNet architecture, a fully convolutional model designed to improve low-light photos by gathering rich contextual information at various scales while maintaining small features. Along with MIRNet, the effectiveness of Convolutional Neural Networks (CNNs) and Generative Adversarial Networks (GANs) for low-light picture restoration are performed using python in pycharm. The efficacy of these methods will be evaluated by means of comprehensive experimentation and evaluation with established criteria for image quality. This study's findings show potential for developing low-light photography in domains such as surveillance, medical imaging, and astronomy, with practical implications for improving image quality in tough lighting circumstances. This work offers a viable solution to the problems related to improving low-light images. Through the integration of sophisticated deep learning architectures and the utilization of carefully selected datasets, to achieve notable enhancements in image quality, especially in low-light situations.