Image inpainting is a hot topic in the field of computer vision. It is a process that enables filling in damaged regions with alternative contents by estimating the relevant information either from surrounding areas or external data. With the advent of big data, image inpainting methods based on deep learning have attracted significant attention in image processing because of their excellent performance. This paper presents a brief review of existing image inpainting approaches and discusses the network structure and performance of each algorithm, along with a comparison of widely used datasets. In view of the existing challenges in this field, this paper proposes potential research directions and developmental trends in image inpainting.