Optimization Of SVC Video Coding Algorithm And Research Of SVC Application

With the development of computer technology and network technology, videocoding technology has been also continuously growing in recent years, at the same time,the fast development of video services for network applications are widely used invarious fields of social life. Currently，the target of video coding technology is not onlyto pursue higher compression efficiency, but also to enhance the adaptability of networkenvironment and terminal devices. Therefore, the Salable Video Coding (SVC)technology came into being, especially in2007, the JVT organization draft H.264Scalable Video Coding standard, which is one of the best solutions of solving thediversity of network environment and terminal device of SVC technology. The videosequence only need be coded once by SVC encoder, then it can obtain video streamcontaining temporal、spatial and quality scalable information. According to the differentnetwork environment and the capacity of terminal devices, users can choose the mostappropriate stream to transport、decode and playback.Firstly, on the basis of study on video parallel encoding algorithm, we have aparallel feasibility analysis on SVC GOP-level, slice-level and frame-level. Then, weuse shared memory multi-thread parallel programming language OpenMP in SVCreference software JSVM to complete GOP-level encoding algorithms. Then，the thesisuse video acceleration instruction of Intel multimedia extensions stream set to parallellyoptimize SVC time-consuming module. The experimental results show that GOP-levelparallel algorithm and the combination of multimedia instruction parallel optimizationalgorithm and GOP-level parallel algorithm, both of them can improve the encodingspeed of SVC obviously.Secondly, the thesis analyzes SVC mode decision algorithm function modules andproposes a fast mode decision algorithm of SVC enhancement layer. Based on the modestatistical distribution correlation between the enhancement layer and base layer, thealgorithm makes use of spatial correlation and temporal correlation betweenneighboring blocks, thus it reduces the candidate modes. Experimental results show thatthe fast mode decision algorithm greatly reduce encoding time with negligible loss of quality and bitrate increment.Finally, the paper presents a SVC video transport application system based on IPnetwork. The program streams are encoded by the optimized SVC encoder, and useMPEG-2transport stream to package. Finally they are packed into IP packets sent to theIP network through TCP mode. The system consists of SVC media server and SVCclient player, it can achieve single screen playback and up to32-way screen browsingplayback. The system streams can achieve adaptive inter-layer switch function. Testresults show that the system is working properly and can achieve good playback resultsin single screen and32-screen. The system can also work in dynamical broadband IPnetwork environment, not only the client can uninterrupted playback, but also the videoquality switch smoothly in this environment. The system makes full use of SVCscalable characteristics.