Advances in Digital Image Inpainting Algorithms Based on Deep Learning

Digital image inpainting refers to the technology of restoring image missing information to automatically repair broken images, widely used in cultural relics restoration, image defogging, film special effects generation. In recent years, the missing information estimation problem of image inpainting can be regarded as a conditional image generation problem with deep learning. Many new ideas for image inpainting come forth. Image inpainting based on Deep learning has become one of the research hot topics in computer vision. This paper summarizes the latest developments of digital image inpainting based on deep learning, and introduces the progress in both convolution methods and network structures. Finally, future research directions are discussed.