| Stereoscopic video can bring a realistic and shocking feeling for users, so people like it very much.Stereoscopic video has incomparable advantages of two-dimensional video, having a wide application inthe field of3D video conferencing, distance learning, stereoscopic video TV, bio-medical, aerospace and soon. But the data of stereoscopic video is twice of two-dimensional video. What’s worse, the realistic andheterogeneous network is dynamic, stereoscopic video packet is easy to be lost, latency and face otherphenomena during the transmission process. Thus it can cause the network transmission distortion. So howdo you measure the distortion of video transmission and achieve data transmission effectively is worthystudying. To solve this problem, this paper unite content factor, proposing a stereoscopic video transmissiondistortion model.First, this paper describe the network video transmission technologies, such as encoding technology,transmission protocol and transmission method. Then introducing common video transmission distortionmodel, including recursive transmission distortion model that based on pixel domain, transmissiondistortion model that based on modelling and transmission distortion model that based on weight estimation,also analyze their advantage and disadvantage. What’s more, we are considering the impact of imagecontent to the model that based on weight estimation to build a transmission distortion model, whichincluding error concealment distortion and error diffusion distortion. In order to simplify the complexity ofthe model, the exponential model is chosen to represent error diffusion distortion. The model considers thecoding mode, the motion vector, the dispartiy vector, the ratio of inter-view coding, also the ratio of intraand inter coding. With the model, we can estimate the transmission distortion accurately at encoding side.Second, this paper calculate the distortion of single view video and stereo video distortion respectivelythat based on the distortion model to get the whole distortion of each frame. Then it uses normalizedmethod to distinguish the importance weight of frame. What’s more, different length of GOP is consideredin the stereoscopic video. Meanwhile, it designs experiment to verify the accuracy of distinction, theexperimental data using the same normalized method. Whether single view or stereoscopic video, from theexperimental result, we can find that the model value is close to the experimental value.Finally, this paper build transmission environment by NS2and Otcl script to change networktransmission condition. Then the frame that marked with the importance is input to the network simulationenvironment. Then combine with the distinguish result of frame importance, we propose an adaptiveactively dropping frame transmission strategy to detect the available network. When the network iscongested, it discard the lowest important frame actively. In addition, a comparative experiment is designedto compare the effect of introducing the self-adaptation strategies. Experiment show that the adaptivestrategies can ease network congestion and improve the perceived quality of the video. |
PDF Full Text Request