Scene geometry can be represented by depth images,which need to be transmitted to 3DTV terminals to assist in rendering of virtual view at an arbitrary view-point. Efficient compression is needed to reduce the expenditure for transmitting depth images. Similarity of motion information between video and depth images is analyzed,and a joint video-depth predictive coding scheme is proposed to reuse the motion information obtained in video coding to assist depth coding. The coding scheme comprises two parts,video-depth motion information copy and video-depth motion information prediction. Experimental results show that the proposed coding scheme can utilize the motion information of encoded video efficiently,and improve the coding efficiency of depth images significantly.