In this thesis, we propose a prediction scheme for stereo video with less computation complexity. The correspondence between disparity and motion vector is exploited and we can reduce the prediction complexity according to this relation. As the compatibility consideration with stereo video system, our prediction architecture is based on the H.264 standard. Simulation results show that the proposed coding algorithm can reduce huge computation complexity and maintain acceptable video quality and depth perception. It needs main and auxiliary images for stereo perception of human visual system, so double data transmission, computation complexity, and hardware resource are necessary for stereo video. Fortunately, we can utilize the relation of stereo image pair to reduce stereo video coding overhead. Because the computational complexity of disparity estimation is much less than motion estimation, the prediction architecture of auxiliary channel can be simplified enormously. Fewer hardware complexity and lower power consumption can be achieved using proposed motion and disparity hybrid prediction architecture. The motion vectors in the auxiliary channel acquired by the motion and disparity hybrid prediction can be used to reconstruct auxiliary frame directly. It means that we can use conventional H.264 decoder to decoder auxiliary sequence without any hardware modification and this ensures the compatibility of usual decoding system.