Research Of Block Motion Estimation Algorithm In Video Compression

With the fast development of multimedia techniques, video compression technology has become a hot spot. As one of the key technologies in video coding, motion estimation deeply affects the speed and quality of video coding compression.This paper first introduces the video compression related concepts and the actuality in study of motion estimation. Then research is focused on matching calculation and search algorithms, which could decide the complexity in motion estimation. For the deficiency of current matching criterions, such as SAD, MSE, this paper proposes a new criterion, called SS_SAD, which utilizes the resemblance of two block's surfaces in matching judging. Compared to SAD and MSE, the new criterion has higher reconstructed image PSNR. An adaptive static-block detection criterion is proposed in this paper, which utilized information in neighbouring frame and blocks to set a threshold value actively, which is used for static block detection. Compared to a fixed threshold value, the new criterion could notablely decrease the miss-judge rate of static blocks. Based on current study of motion vectors statistic results, this paper brings out motion vector horizontal-biased central distribution model and motion vector correlation model. Based on the new models, two fast search algorithms are proposed, which could improve the speed of motion estimation.