Abstract: Automatic road extraction is one of the most researched fields in object recognition from remote sensing images. It is of high importance to the development of geography information technology to achieve automation, intelligence, robustness, and accuracy in road extraction. Base on the physical properties and functions, there are four road image features: the spectral feature, the geometric feature, the topological feature, and the contextual feature. The major road extraction methods are summarized in this paper according to the four features. Their functions and roles in road extraction of the four features are emphasized. The evaluation method for road extraction is also introduced along with the popular road extraction software and sources of remote sensing image. The trend of the research is outlooked at the end of this paper.