Abstract
In recent years, deep learning has become a popular approach for image restoration tasks such as de-noising, super-resolution, and inpainting. In this paper, the current state of the art in applying deep learning techniques to solve these problems was reviewed. In this paper various network architectures and loss functions used in the field are covered, and their strengths and limitations are analyzed. Also, an overview of the evaluation metrics used to assess the performance of these algorithms is provided. Furthermore, several real-world applications were discussed and how deep learning methods have been successfully used to improve the quality of degraded images was demonstrated. Overall, this paper provides a comprehensive survey of the application of deep learning in the field of image restoration and highlights the potential for future research and development in this area.