Tensor Based Consistency Optimization Technology For Streaming Video Transcoding

With the increasing demand for video playback,various video service products have emerged online.Video transcoding technology has a wide range of applications in these products,which is used to solve the contradiction between the original video and playback requirements regarding bit rate,encoding method,packaging format,and other aspects.In recent years,the popularity of high-definition video live streaming has led to new features of large data volume and vital timeliness in video services.Video transcoding has also developed from batch processing mode to distributed computing and streaming computing.However,the development of video transcoding technology has caused network uncertainty,resulting in inconsistencies that can cause additional frame distortion during the transcoding process.This thesis focuses on the research of consistency optimization technology in streaming video transcoding and mainly focuses on the following three aspects of work:(1)A consistency optimization algorithm using tensor technology is proposed to address the inconsistency issue in streaming video transcoding,combining error prevention and recovery.On the one hand,through the primary consistency mechanism of Qt multithreading and the Paxos algorithm,inconsistencies that occur during the transcoding process can be minimized.On the other hand,inpainting technology based on tensor completion is used to reduce the impact of unavoidable inconsistency.The experimental results show that using the method proposed in this thesis can improve the three typical consistency indicators of the target video,effectively reducing the inconsistency issues generated during video transcoding using streaming computing mode.(2)A comprehensive optimization technique based on CAP constraints is proposed for complex scenarios where the cost of availability and partition tolerance cannot be ignored during consistency optimization.Firstly,according to the eventual consistency model used,the consistency expression of the tensorized video copy is given,and the constraint relationship between consistency and availability and partition fault tolerance is derived.Then,different parameters in the streaming video transcoding system are classified and combined with constraint functions to form a three-objective optimization problem.Finally,the NSGA-II multi-objective optimization algorithm solves the optimal allocation scheme set of system resources.The experimental results show that the algorithm can accurately depict the Pareto frontier of the three objective optimization problems when the system parameters are within a reasonable range.The method proposed in this thesis quantitatively analyzes the CAP problem,which is no longer limited to traditional qualitative analysis,allowing for consistent optimization of video transcoding while also considering various needs.(3)A video stream transcoding prototype system was designed and implemented based on the above two consistency optimization techniques.The system adopts a modular design,allowing users to adjust settings to meet different transcoding needs.In addition,the system also comes with consistency indicator calculation and comparison functions,making it convenient for maintenance personnel to evaluate the effectiveness of consistency optimization technology.In testing,the system was able to complete the transcoding and consistency optimization of recorded videos,verifying the effectiveness and practical application value of the method proposed in this thesis.