| As the development of computer,communication and network technology,real-time 2D video visual communication has be come mature,and all video Products have already entered into our lift,study and work,and soon. However,with the requirements of high definition and realistic video quality,tradi1ional 2D video is no longer entirely satisfying the requirements.Multi-view video,which can Provide viewers with the benefits of added realism,selective viewing,and improved scene understanding,Multi-view video coding (MVC) has been proposed for the new generation of the interactive multimedia, and it has been recognized that MVC is a key technology that serves a wide variety of applications,including free-viewpoint television, three-dimensional television and surveillance.etc,and will be the solution to the compression,storage and transmission.etc for 3D interactive video.However a large amount of data is one major obstacle for using multi-view video is the large amount of data.A multi-fold increase in bandwidth over the existing single-view makes it extremely tough to transmit and store multi-view video data.This thesis mainly concerns the problems of highly efficient MVC.To achieve high compression efficiency,co-relation between the different views must be exploited in MVC scheme.We will primary study those aspects of Multi-view Video Coding,such as motion estimation,disparity estimation and Multi-frame.Firstly,we discussed some theory and technology correlative to MVC,Secondly,we study multi-view video coding methods using the H.264/AVC framework.Thirdly,we accomplish the two coding algorithm of MVC on JMVC platform.The first algorithm was proposed a novel disparity estimation method of multi-view video coding based on rectified epipolar lines.we employ the theory of epipolar geometry to rectify the epipolar lines of two views,so that the search range can be confined to a horizontal line with the same y coordinate for any referred point.A table is constructed in our paper to store the relationship between the original image and the rectified one.We do the disparity estimation by simply checking the table.It can efficiently confine the search range in the process of disparity estimation and reduce the total encoding time by 30%~40% with little reduce of PSNR.The method is implemented on JMVC (Joint Multi-view Video Coding).It is the reference Model of multi-view video coding of JVT (Joint Video Team).At present,for finding the best reference frame in MVC,it need to search for all mode.Thus,The second algorithm proposed a novel multi-frame election algorithm for early termination based on H.264.According to probability theory,it can stop search in the average of Rdcost nearby,The proposed algorithm can reduce the computational complexity by early termination multiple ference frames election. Experimental results show that the algorithm can decrease the encoding time by 90%~80% while maintaining nearly unchanged PSNR of pictures and bit rate compared to the FS(full-search) algorithm and TZsearch(TZ-fast-search)algorithm in H.264 reference software JMC4.0. |
PDF Full Text Request