Research On The Predictive Encoding Technology In Video Conference System

Recently, with the rapid development of multimedia technology, video conference system plays a more and more important part in daily lives. Due to its mobility and visibility, video conference system has been paid more attention to. Some key technologies, such as predictive encoding, have been researched as hot topics. At present, the three major operators in China have established IP-based business networks which cover most large cities. The networks have primarily been used to support the service of video conference and video telephone.In this paper, having analysis the existing video conference network of one of the three major operators, only H.323-baesd video conference system is allowed in the existing network, which cannot support the 3G and SIP-based terminal. In order to solve the problem and achieve the communication between the newer and original H.323-based terminal, a multi-terminal video conference platform design plan has been proposed based on the existing equipment. The design of video conference system architecture has been introduced from several parts, such as network architecture, system architecture, terminal units accessing and communicating.As soon as the 3G comes, a more powerful business hosting environment is provided. However, what faced us is a more difficult problem, which is the contradiction of high need of multimedia business and weak compute capacity of 3G mobile unit. In order to improve the quality of video and reduce the computation of terminal unit, cutting encoding time and increase encoding efficiency is in an urgent need. Having made deeply research on H.264 protocol, motion estimation is one of the most important factors which can impact on the video compression efficiency. Be aimed at the disadvantages of UMHexagonS algorithm used in H.264, the author proposed a more excellent algorithm named Self-adaptive Cross Multi-hexagon Shaped (SCMS). Compared with UMHexagonS algorithm, redundant search points are cut in SCMS and self-adaptive search templates are introduced in order to reduce the number of search points.At last, the SCMS algorithm is validated by various standard test cases. What is same as prediction is the result, which shows that, as a prerequisite of ensuring the SNR unchanged, the number of search point has been cut and complexity has been reduced in optimized algorithm. Much time can be saved in SCMS motion estimation process, which results in improving the real-time characteristic of video. The SCMS algorithm is worth in high real-time scene, such as the video conference system in this paper.