| Along with the popularization of computer and network technology in the world, people need more and more communications. The content of communications expand from voice and texts to video. So the multi-media technology have been the hotspot in the information and communication area, and the coding, storage, processing of video is the core area of the multimedia. Since long time, transform-based data compressed technology always take leading position in video coding application field. Based on dispersed cosine transform video coding method is adopted in all of international standards for video coding, such as ISO's MPEG-1, MPEG-2, MPEG-4 video compression standards, and ITU-T's H.26X which is a series of video coding suggestions. For excavate the potential of video data encoding, this dissertation is based on the theory of transform coding, study the method of transform coding of video compressed for improving efficiency, and we have developed two different video coding methods based on wavelet transform and matching pursuit, respectively. The result of experiment shows these coding methods improve coding efficiency on different levels compared with H.263 and H.26L. These methods have also established experiment emulation platform for the later research of this dissertation. Besides the research of transform coding method, the study of rate-distortion analysis and optimization method also plays an important role in the video coding and communication system. Its purpose is to optimizate rate-distortion for video coding in the transmission process, so that the best visual quality can be gotten under limited transmission condition. In the traditional rate-distortion analysis method, rate and distortion is about the function of quantification parameter. Because it has very strong nonlinear features, its calculation is very complex and inaccurate. Plenty of experiment datas are analyzed, and the concept of transform-rate is presented in this dissertation. Based on the concept of transform-rate, we present a new rate-distortion model in which rate and distorted functions are about the functions of transform-rate parameter. The rate function has good linear features, and the distortion function depent little on source. With these characteristics, rate-distortion optimization for transform-based video coding may be realized simply, fast, accurately, and robustly. In order to apply this rate distorted model extensively in the video coding system, we have presented general rate-distortion curve prediction algorithm, rate control algorithm, bit distribution optimization algorithm and rate shape smoothing algorithm for transform-based video coding based on rate-distortion model with transform-rate, and have carried out extensive experiment in three difference video coding systems. The accuracy of presented rate-distortion model and the robustness of corresponding algorithm have been verified by experiment results. In video coding data transmission, more corrupt image quality may be caused at decoder because of error and packet loss, etc. For realize robustly data transmission, the end-to-end distortion of video transmission is analyzed detailly in this dissertation. With the analysis, we have presented a fast prediction algorithm for end-to-end distortion that is used in joint source-channel rate-distortion analysis. Combining with the research of the method for coding model selecting, we have presented a fast rate-distortion optimization algorithm. Experiment result shows, the end-to-end distorted prediction algorithm can predict channel error accurately, and the rate-distortion optimization algorithm has improved video quality of end-to-end video coding transmission efficiently. As a project of typical application, we have developed TSM4 video encoder based on the MPEG-4 VM8 for "Easy Study" system that satisfy the requirement of the teaching of real time. This paper has discussed the realization and optimization method of TSM4 video encoder in detail. Experiment result shows the optimizated TSM4 video encoder can realize real-time encoding for two CIF videos on the general processor of Intel Pentium III 800.
|
PDF Full Text Request