Image enhancement has been a very significant part of computer vision. The image captured in low light has a poor quality and includes a lot of noise which has a negative impact on various learning models. In the past decade, Artificial Intelligence technology has advanced more rapidly than ever, and a wide range of its applications, including autonomous vehicles, AR, VR, speech recognition, picture identification demands enhancement of images. Currently, the available methods give promising results with optimally lit images but have a poor performance against low-light images. The image captured in low light has a poor quality and includes a lot of noise which degrades the performance of vision-based algorithms. To make the details buried in the image more prominent, reducing the noise and blur from the image is crucial. The advancements in deep learning have introduced numerous techniques to improve the image quality under dim light. A great deal of research has been devoted to this in the past to boost the quality of images clicked in low light conditions. Numerous deep learning-based approaches are proposed to solve the issue. This paper presents a thorough survey of deep learning-based models and improves an attention-based convolution network - Attention U-Net aimed at improving image perception and interpretability in an environment with poor illumination.