An efficient motion compensation algorithm based on double reference frame method

Recently, several techniques based on the multiple reference frame scheme (Proceedings of the Dada Compression Conference, March 1998, pp. 239-248; Proceedings of the European Signal Processing Conference, September 1998, pp. 253-256; Proceedings of the International Symposium on Information Theory, August 1998, pp. 409-418; IEEE Trans. Circuits Systems Video Technol. 9(1) (February 1999) 70-84; Proceedings of the IEEE International Conference on Image Processing, October 1997, pp. 17-20) have been proposed to improve the motion prediction gain. Though these techniques yield higher prediction gain than the single reference frame scheme, they cannot maintain acceptable image quality at very low bitrate. Besides, the blocking artifacts may be visible along the block boundaries, since each block is predicted independently of its neighbors. And they require tremendous computational complexity for the motion search. To overcome these drawbacks, this paper proposes a novel motion compensation algorithm, based on the double reference frame, the double motion vector, and the searching position shifting (SPS) schemes. First, to reduce the motion vector bitrate and the computational complexity of motion search, we constrain the number of reference frames to 2, and use only two motion vectors per block. Second, to alleviate the blocking artifacts and to yield better prediction performance, the SPS scheme is introduced. Experimental results demonstrate that the proposed algorithm yields about 3-4 dB higher prediction gain than the single reference frame scheme. Subjective quality is also improved by alleviating the blocking artifacts. In addition, the proposed algorithm can be combined with a robust video transmission scheme. Thus, even if the channel experiences errors during the transmission, the proposed algorithm can efficiently protect the image quality.