Research On The Motion Estimation Based On Multi-view Video Coding

As an advanced video coding standard, H.264/AVC offers a very good video quality. In the coding of video compression, H.264/AVC increases obviously the coding quality and compression ratio. As network-friendly, the coding efficiency can be enhanced by 50% than H.263 and MPEG-4 in the same subjective quality. Multi-view video coding (MVC) is the fourth amendment of the H.264/AVC.Multi-view video is captured by synchronized cameras from different viewpoints. This coding method increases scene depth and characterization of information. Comprised of rich 3-D information, this technique is capable of rendering realistic natural contents into the video sequences. Thus it has become a hotspot of video research. However, Multi-view video results in a tremendous amount of data from numbers of cameras. How to effectively encode multi-view video sequences is a constraint to the application of multi-view video. The motion estimation consumes the most time of the total multi-view coding. Consequently, the key problem of the MVC is how to effectively estimate the movement, reduce the coding time-consumption and increase the coding efficiency.In this paper, an adaptive early termination algorithm for motion estimation in MVC is proposed to meanwhile maintain good video quality. Firstly, analyses of some traditional coding algorithms and their shortcomings are given in this paper. Secondly, a method of setting the adaptive threshold based on each group of pictures (GOP) is proposed. Finally, the adaptive early termination algorithm for motion estimation is formulated. The algorithm is implemented in JMVC (Joint Multi-view Video Coding), and experimental results demonstrate that the time-consumption in video coding reduces obviously with negligible video-quality degradation. The algorithm increases the coding efficiently and robustly.