Research On Technologies Of Video Stream Multichannel Realtime Compression And Transmission : Improvement And Application

In proportion as the popularization of smart terminals spreads the advance of online video applications of the Internet intensifies. Under the constraints of limited storage space and network bandwidth, the importance of video compression and transmission technologies arises. Researches on signal characteristics, performance measures, prediction models and coding strategies, contribute to robustness of multimedia storage and transmission facilities, possess certain theoretical and practical significance.Based on researches on modern video coding standards and technologies, basic video coding methods are systematically introduced, fundamental mechanisms under video compression and transmission systems are explained and implemented. Based on H.264 reference software, to improve decoding real-time capabilities, memory management techniques of decoding process are designed, and a real-time video analysis software is implemented. And to improve encoding real-time capabilities, two real-time optimization techniques are investigated and implemented: 1. Human Visual System based real-time quantisation techniques and 2. Rate Control techniques for Joint Source Channel Coding. The contributions of this paper are outlined below:Information coding theory and statistical characteristics. Related information coding theory, statistical characteristics, and coding methods are introduced, which including concepts of coding efficiency, four specific coding methods, and video signal statistical characteristics.Key techniques used in video compression and transmission systems. A system of video compression and transmission is designed and implemented, and similar real-life systems utilized key techniques are presented, which including message pump realized multi-thread cooperation, multi-platform video signal capture methods, original video signal formats, open-source softwares based codec operation programs, multi-media real-time transmission protocol, and multi-level display buffer management.Central coding methods used in modern video coding standards and memory management for decoding process. Respecting to modern video coding standard H.264, whose advanced coding methods achieve fairly good efficiency, the central coding methods in H.264: prediction and transformation codings are concretely expounded. The baseline decoding process is concretely integrated. And improvement schemes of memory management of decoding process are implemented: 1. Static memory based dynamic management technique and 2. Virtual memory management technique.Human Visual System based real-time quantisation techniques. Two real-time quantisation techniques are investigated: 1. Luma and chroma eyes sensibility inequivalence based quantisation, and 2. Psychovisual enhancement quantisation. Results demonstrate that HVS based real-time quantisation techniques enable exchange of PSNR and rate, or improve subjective visual quality by reasonable distribution of bits.Rate Control techniques of Joint Source Channel Coding. Methods to build R-Q and R-D models using Q-R-D data fitting are demonstrated. Hyperbolic, logarithmic and hyperbolic-logarithmic functions fitted curves are investigated and compared, a concrete models building method is presented. Experimental results demonstrate that weighted sum of hyperbolic and logarithmic functions achieves better accuracy than either. Building models using few amounts of data is also researched, experimental results demonstrate that using few amounts of data is able to achieve fine result, and better accuracy can be achieved while more data are provided.