This paper proposes a video information-embedding scheme in which the embedded information is reconstructed without knowing the original host video. The proposed method presents high rate of information embedding and is robust to motion compensated coding, such as MPEG-II and MPEG-III. The characteristic of the proposed scheme is to use pixels contractive relation to assist motion compensated coding process. Information is embedded in video frames using the block DCT. Embedded frames are then coded using MPEG-III. After that the information will be extracted from video with high efficiency. Experimental results have indicated that the method is robust against MPEG encoding and re-encoding.