| With the rapid development of VR(Virtual Reality)technology,panoramic video gradually enters people’s field of vision,and brings people an unprecedented and immersive experience.Compared with traditional video,panoramic video has the characteristics of high resolution and high bit rate,and the rapid increase in data volume brings huge challenges to the transmission of panoramic video.The tile-based dynamic adaptive transmission scheme of panoramic video transmits high-quality video within the user’s field of view and low-quality video outside the field of view by dividing tiles,effectively reducing the demand for bandwidth and decoding resources.For the panoramic video of CMP(Cubemap Projection)format,this thesis implements a tile-based dynamic adaptive transmission system based,and studies the user experience quality evaluation model and adaptive download strategy.Aiming at panoramic video of the efficient CMP format,this thesis designs and implements a tile-based panoramic video-on-demand system.The server module completes tile division,encoding,and media presentation description file generation according to the DASH(Dynamic Adaptive Streaming over HTTP)protocol;the client is based on a standalone head-mounted display to complete the download,decapsulation,decoding and rendering of panoramic video.Experimental results show that the system can provide users with a good panoramic video viewing experience when bandwidth and hardware conditions are limited.In a tile-based panoramic video-on-demand system,the perspective change will result in lower video quality in some areas of the field of view.For this kind of panoramic video system with priority viewing angle,this thesis systematically studies the user experience quality evaluation model through subjective experiments.First,for the situation where the video quality is consistent in the field of view,based on subjective experiments on the VR stand-alone head-mounted display,an evaluation model for the resolution,bit rate and experience quality of the panoramic video is established.Then considering the low-quality situation in some areas in the field of view,this thesis analyzes the influencing factors such as the ratio of high-low-quality areas and low-quality areas,and establishes a user experience quality evaluation model.The experimental results show that the objective quality predicted by the model proposed in this thesis and subjective quality are in good agreement,and the real-time experience quality of users in the on-demand system can be predicted more accurately.Tile-based panoramic video-on-demand system not only needs to adapt to channel bandwidth fluctuations,but also needs to consider the impact of viewing angle prediction errors.Based on the established user experience quality evaluation model,this thesis proposes a rate adaptive download strategy based on MPC(Model Predictive Control).Based on parameters such as prediction bandwidth,viewing angle prediction accuracy,the rate of video base layer,the rate of enhancement layer,etc.,with the goal of maximizing user experience quality,this strategy adaptively determines the rate of the tile to be downloaded.This thesis further analyzes the impact of the viewing angle prediction accuracy and the completion probability of tile download on the proposed download strategy.Experimental results show that the rate-adaptive download strategy based on MPC can better realize the rate adaptation of the video in the user’s field of view,and effectively improve the quality of user experience. |
