Research Of Multi-screen Video Content Porcessing And Transmission

With the development of network-based transmission technology,people can view the video at various terminals.Video resolution is becoming higher nowadays.Also,Concept of HD,UHD and even 8K comes one after another.According to Cisco's forecast,thanks to the explosive growth of mobile terminal,video content will occupy 70% or more of the throughput data traffic in the future,and this proportion will continue to increase in the next few years.Video processing and transmission will become a huge problem.Due to the cloud's compatibility of computing resources and storage resources,cloud-based video processing has gradually become a hot topic.As an open source project,Hadoop distributed computing framework has been the hot spot since itpublished.However,framework based on Hadoop has two problems.First of all,Hadoop does not support the iterative calculation;Second of all,Hadoop based framework can't cope with non-equivalent computing nodes.In this paper,in order to solve these two problems,we propose a Celery based videos distributed computing framework,which is designed to be compatible with larger computing resources,and efficient with a large cluster at the same time.In this paper,we proposed a distributed video processing system based on celery.Using Hadoop's structure as a reference,we propose a three-tier structure,including API server layer,JobTracker layer and TaskTracker layer.This structure enables our system to be compatible with both computing resource and functional extension.In this paper,we present the structure in detail and how we optimize it.The experimental data indicate that the system is efficient with multiple calculating nodes.Video streaming in mobile devices has always been a hard problem,and MPEG-DASH is considered to be a solution.DASH protocol requires the server to offer multiple profiles at the same time.We can get a smoother playback as the profile number grows.However,we notice that the increase of profiles will bring in excessively frequent switches which in turn cause a QoE loss.Besides,growing numbers of media segments will put large burden on the backbone network.In this paper we use a proxy-assistant architecture to address above two problems.At the client side,we propose a new rate adaption method to ensure a smooth playback at the expense of a little average overall bandwidth.On the other hand,the smart proxy tries to aggregate similar requests with an importance map.Compared with prior proxy based scheme,our scheme is proved to aggregate alike requests to reduce promote cache hit ratio.The result shows that our method reduces the switch times and server requests times with a little overall average bitrate loss,and consequently make our scheme both QoE-friendly and efficient under low-delay and multiple profiles scenario.