The Research On Fast Algorithm Of Multi-view Video Coding Based On H.264

As a link of video processing, video coding technology provides foundation forvideo transmission and display. Video coding technology has developed from H.261to H.264, from2D to3D,continuously developing into more mature technology. As akey technology of3DTV(3D Television),the multi-view video coding technology getsmore and more attention from experts and scholars, and becomes a research focus inthe field of video. Multi-view video contains a set of camera from many angles forthe image sequence, and there are high correlations between them. The sense ofmulti-view video is more vivid, and the stereo feeling is more direct. Meanwhile theamount of data of stereo video is greater than it of traditional video. For transmissionand storage, great amount of data is a huge challenge. Therefore, the multi-view videocoding technology should be stronger in data compression and more efficient incoding, namely compress the video efficiently.The thesis analyzes and researches multi-view video coding technology based onmulti-view video coding model JMVC(Joint Multi-view Video Coding),then putforward optimization algorithm in view of mode decision and fast search algorithm:firstly, analyze the fast search algorithm TZSearch in JMVC, and present a new fastsearch algorithm. The new algorithm combines the fast search algorithm SimplifiedUMHexagonS(Simplified unsymmetrical cross Multi-Hexagon search) used intraditional single view video coding with TZSearch. The proposed algorithm usesdifferent search algorithms for inter view prediction and inter frame predictionrespectively to reduce search time and improve search efficiency. Secondly, analyzethe relationship between the mode of current macroblock and the mode of adjacentcoded macroblocks and the corresponding position of reference frame. According tothe relationship between the modes of macroblocks, optimize the mode decisionalgorithm of current coding macroblock, and reduce the division and search of refined small block.The improved fast search algorithm is effective to the sequences containing fastmotion and big motion area，and the improved mode decision algorithm is effective tothe sequences containing slow motion. Therefore, we combine the two improvedalgorithms, and apply them to10sequences of different characteristics. Theexperimental results show that the improved algorithms get significant effect.Compared with the original algorithm in JMVC, the improved algorithm can reducethe encoding time by83%under the premise of insuring reconstruction video qualityand coding overhead.