| The panoramic video can record the image of 360-degree field of view,and the audience can freely switch the field of view to obtain an immersive experience.However,streaming the complete panoramic video with full resolution requires a high bandwidth network,and audience’s field of view only uses part of the transmitted image,so a large amount of transmitted data is wasted,resulting in low transmission efficiency.Tile-based panoramic video streaming only transmits the tiles needed by the audience’s field of view,which has the advantages of high transmission efficiency and high field of view quality.However,since the field of view is composed of multiple tiles,it is necessary to deal with the bit rate allocation and video buffer management of multiple video streams of tiles.At the same time,it also needs to deal with the impact of network fluctuations and field of view switching on the quality of experience(QoE).The existing methods mainly realize QoE guarantee from the aspects of field of view prediction,transmission redundancy tiles and optimization of comprehensive QoE.But they do not consider the guarantee of transmission efficiency and the different requirements for tiling schema in video tiling process of different panoramic videos,resulting the problem of transmission efficiency degradation in the process of the poor transmission control and unreasonable video tiling schema.This paper aims to study how to optimize the transmission efficiency on the basis of guarantee QoE,so as to achieve an efficient and stable highquality panoramic video block transmission service.Firstly,for the process of transmission control,this paper designs a transmission control model to ensure transmission efficiency based on constrained QoE.Deep reinforcement learning technology is used to comprehensively analyze multiple QoE indicators and transmission efficiency to realize bit rate allocation and video buffer management.In addition,aggregation transmission control and two-stage decision-making mechanism are proposed to uniformly control the tiles according to their category,and the unstable factors such as network bandwidth and field of view are unbound from the core decision-making stage,so as to optimize the complexity and stability of the algorithm.Secondly,in the process of video tiling processing,this paper proposes a dynamic adjustment framework of video tiling parameters based on MAPE-K model.It collects multi-user historical viewing logs and defines multiple aggregation analysis indicators to identify the characteristics of viewing behavior.The tiling health indicator is also defined to monitor whether the video need to be processed with new parameters.When the health is lower than the threshold,the tiling parameters will be adjusted to finally generate tiled video resources which are more suitable for the transmission of this panoramic video.Finally,this paper implements a tile-based panoramic video transmission system,and tests the system functions and the performance of proposed methods.Experimental results show that the proposed method can effectively ensure QoE and transmission efficiency in tile-based panoramic video transmission. |
PDF Full Text Request