In the field of computer vision and photogrammetry, image dense matching has always been one of the key and difficult point of research. It plays an indispensable role in the three-dimentional reconstruction and digital surface model production. In recent years, with the rapid development of the deep learning, image dense matching algorithms based on deep learning have also emerged endlessly. This article reviews the current representative technical methods, and describes the research progress and trends of deep learning in image-dense matching based on two types of research ideas: image block-based similarity measurement learning and end-to-end disparity map generation. The advantages and disadvantages of the existing methods are presented in order to provide a literature review with reference value for the research of image matching.