Reasearch On The Transmission And Improvement Technology Of Multi-view Auto-3D Video Over Large-scale Networks

Vision is the most efficient way for human to acquire information. Because thescenes people see in nature are3D, people always dream of rebuilding the3D scenesby image. Multi-view auto-3D video is a milestone of3D video technology, whichmakes the viewers see3D image without eye equipments. The development ofbroadband technology makes a large amount of data transmission on the internetpossible. The multi-view auto-3D video based on network will bring people brandnew audio-visual experience and become a new business field of the next generationhigh confidence networks. The multi-view auto-3D video is characterized by high bitrate and low tolerance for bit error, so the large-scale commercial multi-view auto-3Dvideo based on network must make improvements on data transmission, video codingand video play.The work of this paper is as follows: In order to solve the problem of the high bitrate and low tolerance for bit error, making improvements on the data transmission,video coding and video play of the large-scale commercial network multi-viewauto-3D video.In the firs, a P2P peer connection filtration mechanism and a transmissionscheduling mechanism were proposed. The purpose of the mechanism is to improvethe data transmission between clients by providing the users with the peers which islow loss rate and high transmission rate. Experiment results demonstrate that thescheme improves quality of service and the video FPS is higher and more stable. Datatransmission rate increases by15%.In the second, a network multi-view auto-3D video transmission system witherror correction scheme coded by LDPC was proposed. The system employs LDPC tofulfill the error correction caused by the packet loss. With the error correction scheme,the transmission quality increases greatly and solves the low tolerance for bit errorproblem. The experiment results demonstrate that the error correction scheme workswell and the average PSNR of multi-view auto-3D video increases dramatically.In the end, a scheme which links the distributed video coding (DVC) with theCUDA video decoding acceleration mechanism was proposed. The scheme employsDVC to improve the tolerance for bit error and decrease the encoding complexity, employs GPU to do the complicated DVC decoding which release the burden of CPU.The experiment results show that the mechanism is feasible which increases theaverage decoding process dramatically compared to the traditional decoding schemeand the FPS reaches the video play requirement.