| With the continuous development and popularization of 3D technology,stereo video has become an important technology application field.The stereoscopic video refers to the video technology that can present stereoscopic visual effects.It can not only bring more realistic visual effects to the entertainment industry such as movies,TV,and games,but also provide more real and vivid display methods in the fields of medical care,education,science,and so on.It is of great significance to promote the development of various industries and improve people’s quality of life.As a typical representative of stereoscopic video,3D naked-eye video is constructed by multi-channel video stream.The huge data load in the video stream and strict delivery time demand bring severe challenges to high-quality video transmission.First of all,the existing streaming media transmission schemes adopt the ”serverclient” or ”end-to-end” mode for connection establishment and service transmission,which cannot solve the unified delivery of multi-channel concurrent video streams in3 D naked-eye videos.In addition,video terminals require fine-grained and highly accurate bandwidth estimates to determine the optimal send rate in time to ensure low latency and low packet loss rate for service traffic.However,the existing bandwidth estimation algorithms,such as NewReno,Cubic,BBR,etc.,cannot guarantee the accuracy of bandwidth estimation in a small time scale in the network environment with severe fluctuations.Therefore,the main contributions of this paper are as follows:1.Through theoretical and experimental analysis,the problems of low bandwidth utilization rate and slow bandwidth estimation speed in real-time 3D naked eye video transmission are verified.On the one hand,when the multi-channel concurrent video streams of 3D naked-eye video are transmitted in the traditional end-to-end mode,they compete with each other,which leads to the decrease of bandwidth utilization.On the other hand,most bandwidth estimation algorithms converge slowly in the fluctuating network environment,which cannot guarantee the timeliness and accuracy of bandwidth estimation.2.Aiming at the problem that the existing schemes transmit multiple concurrent video streams and compete with each other,this paper proposes an ”Agent-to-Agent”transmission scheme.The main idea of this scheme is to centralize processing,unified transmission and macro management of multi-channel video streams through terminal Agent.In the experimental environment,the ”Agent-to-Agent” transmission scheme can effectively avoid the competition between traffic,so as to achieve the bandwidth utilization of 100%.The proposal of ”Agent-to-Agent” scheme makes up for the shortcomings of the current ”end-to-end” mode and ”server-client” mode when transmitting concurrent data streams,and provides a feasible reference scheme for the ”multiflowto-one” transmission mode.3.In view of the problem of slow convergence caused by the ”iterative” method used to guess the link bandwidths in the Bandwidth Estimation algorithm,this paper proposes a Direct Bandwidth Estimation(DBE)algorithm using ”direct measurement”to estimate the maximum available bandwidths.The available bandwidth is directly detected by the packet interval model and packet rate model,and the results are arbitrated by the rate decision algorithm.The experimental results show that the algorithm can obtain accurate bandwidth estimation in various environments with very small time cost and is suitable for delay sensitive services. |
