Research of Motion Estimation Algorithm in Video Compression
|School||Nanjing University of Posts and Telecommunications|
|Course||Signal and Information Processing|
|Keywords||Video Compression H.264 Motion Estimation UMHexagonS JM10.1|
With the rapid development of multimedia technology, rich video information is an importantmeans of access to information.However, the huge amount of data makes the video storage andtransmission become very difficult, video compression algorithm is the key to solving this problem.H.264is jointly developed by ISO/IEC MPEGand ITU-T VCEG. The standard uses a variablemacroblock, multiple reference frames, and other technology, under the same image quality, itscompression rate has more than doubled compared with H.263.But the improvement of codingefficiency is gotted by increasing computational complexity of the cost, which seriously affected theapplication and promotion of real-time encoding. The motion estimation is the core module in videocompression, its efficiency will directly affect the level of coding efficiency.Therefore, efficient,low complexity and easy to implement motion estimation algorithm to become a research hotspot.This paper first introduces the basic principles of video compression coding and H.264basicstructure, key technologies，then research the classical block matching algorithm according to theprocess of the motion estimation algorithm, gives the advantages and disadvantages of each.Unsymmetrical-cross Multi-Hexagon-grid Search(UMHexagonS)algorithm is used as integer pixelblock motion estimation algorithm in H.264official test software JM (Joint Model). Finally,in-depth analysis of the algorithm, I propose three improvements, they are: The improvement ofstarting search point prediction(change cost function pattern)、Adaptive cross search pattern、Hexagon and diamond pattern modification and so on.In JM10.1model, the C programming language to achieve improved algorithm, and afterVS2008compiled, got simulation results. After testing five sequences with different motioncharacteristics indicates: Compared the improved algorithm with the original algorithm (10reference frame), peak signal-to-noise ratio (PSNR) and bit rate (Bit Rate) close to the case, themotion estimation time average decreases of19.02%, up to28.2%, encoding time is reduced by anaverage of13.36%, up to20.9%.