Research Of H.264Motion Estimation Algorithm

With the increasing number of video cameras, the amount of information surveillancevideo recorded is also growing. When monitoring person need to find suspicious monitoringinformation in a specific location, there are a lot of useless information in surveillance video,and they need to spend a lot of time and manpower. In response to this situation, videosynopsis-a quick surveillance video view summary technology, can simultaneously presentmultiple objects and activities that occur at different times. Research in this area has importantpractical significance.Motion estimation is one of H.264video compression encoding the key technologies. Wemainly focus on the analysis of a variety of information which is obtained by motionestimation algorithm, and apply it to video analysis and storage. The study is expanded fromthe motion estimation algorithm, and make compressed domain video analysis and storage ascore research, and aim at improving moving object extraction accuracy and to theeffectiveness of video synopsis compression, and research on motion vector preprocessing,background modeling, residual modification, data acquisition and conversion, and proposetwo algorithms respectively for two research objectives:(1) We use the compressed domain background modeling method to extract the movingobject quickly. Pixel-domain analysis, the mainstream approach to analyze surveillance video,has always been a hot issue in academy and industry. However, with the increasing volumeand resolution of surveillance video, the flexibility and efficiency of fast processing isgarnering more significance. Under this circumstance, surveillance video analysis in thecompressed domain is indeed of strategic importance from the angle of balancing visualperception and processing speed, especially in modeling background and segmenting movingobjects. Therefore, a compressed domain based scheme is proposed to model background andsegment moving objects in this paper. The main work and achievements are as follows: Firstly,a compressed domain scheme is established to calculate the local binary pattern (LBP) valuewith the motion vector (MV); Secondly, a background modeling method with the MotionVector based Local Binary Pattern (MVLBP) is applied to the compressed domain.Experimental results show that our approach gives a stable performance in shorter time.(2) We presents a video synopsis compression method based on the three-dimensionaldiscrete cosine transform. The separated moving object foreground object pipeline wasgenerated in video synopsis generation process. With this characteristic, we combine with themotion estimation algorithm to re-extract the residual data, and then process the data withthree-dimensional discrete cosine transform, in order to achieve the video synopsiscompression. The experimental results show that the method can reduce the bits’ number ofthe coded stream, and remove the temporal redundancy in the digest video data.Finally, to generate the video summary and a video summary of storage combine tocomplete a simple video analysis and retrieval subsystem.