Deep Learning for Single Image Deblurring

Atraditional assignment in basic computer vision is picture deblurring, which involves recovering a clear image from a hazy input picture. Significant progress has been made in overcoming this problem thanks to developments in deep learning, and numerous deblurring networks have been proposed. To provide the community with a helpful literature review, this work provides an extensive and timely evaluation of newly released deep-learning-based image deblurring techniques. We begin by going over frequent reasons for image blur, introducing performance measurements and benchmark datasets, and then summarising various approaches to the problem. We then provide a thorough analysis and comparison of convolutional neural network (CNN) methods utilizing a taxonomy based on architecture, loss function, and application. Furthermore, we talk about a few domain-specific deblurring applications, such as text, face photos, and stereo image pairs. Finally, we address important issues and potential future research avenues.