Research And Implementation Of Light Field Video Transmission Based On 5G Edge Computing

With the development of the fifth Generation(5G)mobile communication network,its performance advantages will greatly promote the application of immersive media.Light Field(LF)can efficiently and comprehensively capture the three-dimensional information of the scene space,which is expected to become the representative service of 5G immersive media applications.LF transmission requires high bandwidth and low latency,and the task of viewpoint rendering also consumes a lot of computing power.So,it also puts forward high demands for the computing and communication resources of the network.Mobile Edge Computing(MEC)can provide services for terminal devices at the network edge,which effectively meets the needs of LF transmission.Based on this,this dissertation studies the hierarchical transmission of LF data and the optimization of MEC-assisted LF delivery architecture.Firstly,a reliable hierarchical compression scheme is the prerequisite for efficient LF transmission.The traditional light field compression schemes did not consider targeted optimization design based on human visual characteristics,thus failing to achieve the optimal user viewing experience quality and light field data compression transmission efficiency.Therefore,this dissertation analyzes the main service characteristics and data structure of LF video,and an coding scheme based on LF user content awareness and viewpoint synthesis is proposed.By extracting the motion and depth information of each LF frame,a user perception model can be established and the viewpoint frame will be divided into three types of LF Region-Of-Interest(LF-ROI)macroblocks.Besides,an LF Multiple Layers Coding(LF-MLC)based on viewpoint synthesis is designed,which can control the decoding LF quality by adjusting the number of coding layers.Through LF-MLC,the quality hierarchical coding and transmission of macroblock arrays in different LF-ROI levels can be achieved,which can ensure the stable quality of users’ subjective visual experience and minimize the network bandwidth consumption of light field data transmission.Different from the traditional multimedia transmission,the light field hierarchical transmission scheme based on LF-MLC needs to transmit more coded data in the network and execute the viewpoint synthesis algorithm with higher computational complexity to decode the LF content at the receiving end.This dissertation introduces the MEC architecture to assist the hierarchical transmission service of LF,which can not only give full play to the bandwidth advantages of the 5G access network,but also offload the terminal computing to the edge to reduce the service delay of LF transmission.In the LF transmission edge network with multiple edge nodes and multiple users,this dissertation models the viewpoint rendering task consumption based on LF-MLC,discussing the service node access selection,task offloading strategy,and bandwidth/computing resource allocation of the entire network system.An online Joint User Association and Resources Allocation(JUARA)optimization algorithm based on deep reinforcement learning is designed,which can balance the service expenses and response delay of LF service in a long-term time-varying network environment.Simulation results verify the effectiveness of edge-assisted LF transmission architecture and the proposed JUARA algorithm.