| Video coding is one of the key technologies of multimedia communication. Lots of abstracting digital video services is changing the living styles of human rapidly. Video coding technologies give us high quality and high resolution visual enjoy, and also push the development of information globalization. The goals of the video coding were changed from the storage to the transmission. The latest FGS (Fine Granularity Scalability Coding) adapts to network bandwidth variations and tolerates transmitted errors (such as packet losses in the Internet network and random or burst errors in wireless network). Absolutely, FGS is one of the most important technologies; and it has a wide application.Coding Efficiency is the kernel goal of video coding. Although the Internet-oriented FGS coding schemes has good scalability, its coding efficiency is still poor compared to non-scalability schemes. The conventional coding schemes remove the redundancy to get good coding efficiency by all kinds of means. The DCT coefficients of nature image have good energy compaction, so DCT is adopted by most of coding standards. But the coding efficiency of DCT will be low when it does with the motion residuals. The thesis systematically researches the video coding technologies for communications. The main idea of the thesis is that research new method to improve the coding efficiency of DCT, then deal with the problem of poor efficiency of FGS. The main work is as follows:1. Based on the analysis of the essential reason of poor coding efficiency of DCT for motion residual, we thought that much redundancy information still exist in coding stream. Based on analysis of statistical distribution of motion residual DCT coefficients, a hiberachical motion compensation scheme was proposed. The method puts all of the non-zero quantized DCT coefficients of residuals into two groups, one of which are not necessary to code their level information, then the rate can be reduced. On the other hand, the process of setting zero can be regarded as a process of"quantization noise reshaping". The total power of noise is reduced, so the quality of images is improved. Hiberarchical motion compensation scheme improves the coding efficiency by two hands: bitrate and quality. The problem of coding efficiency for DCT is resolved.2. In this thesis, we focus on the accuracy problem of DCT-domain motion compensation, and propose a new block matching criterion—quantized DCT-domain partial coefficients block matching criterion. The criterion keeps the advantage of MSE power matching, and puts the power of quantization noise into matching computation as a parameter. The new criterion ensures the minimum power of reconstructed residuals, so the rate will be the lowest. 3. In this thesis, we focus on the complexity problem of DCT-domain motion compensation. We think that the key part of computation complexity of the DCT-domain schemes is the DCT coefficients restoration of non-aligned blocks. We propose a fast algorithm base of unequal block matching criterion. In this new algorithm, only the power of significant coefficients be calculated. So the total complexity be reduced.4. This thesis analyses the application background and the architecture of Fine Granularity Scalability Coding scheme, and points out its weakness is poor coding efficiency compared to the non-scalability coding schemes. The hiberarchical compensation embedded scheme is proposed. The new scheme improves the efficiency obviously.
|
PDF Full Text Request