The accurate registration of visible image and three-dimensional point cloud is the prerequisite of image fusion. How to improve the registration accuracy is the focus of current research. In this paper, a registration algorithm of visible image and three-dimensional point cloud based on point features is proposed. Firstly, the visible image is preprocessed to correct the distortion of the camera. Then, the appropriate self-build calibration field is set, and the homonymous feature points in visible image and point cloud are obtained by using the blob detection algorithm. Finally, the corresponding registration model parameters of visible image and point cloud are calculated based on multiple pairs of homonymous feature points. Experimental results show that this method can achieve centimeter level registration accuracy.