Image Deblurring Using Deep Learning: Applications, Methods, and Future Directions

Abstract

A basic challenge in low-level computer vision, image deblurring seeks to restore a clear image from an input image that is blurry. With the advent of deep learning, several deblurring networks have been suggested, and much progress has been made in resolving this issue. With the aim of providing a valuable literature review for the community, this article offers a thorough and up-to-date evaluation of deep learning-based image deblurring methods that have recently been published. Following an overview of typical picture blur sources, they provide benchmark datasets, performance measurements, and a summary of various issue formulations. The next stage is to provide a comprehensive study and comparison of convolutional neural network (CNN) approaches, organized according to architecture, loss function, and application. Moreover, they go over a few domain-specific deblurring uses, such as those for face photos, written content, and stereo image pairings. Finally, they go over some important obstacles and potential avenues for further study.