Research And Design Of The Context-Based Arithmetic Coding

With the development of the communication technology in recent years, the high resolusion digital video is used more and more widely, so the research on more efficient video coding technology has been a hot point. Context-based adaptive binary arithmetic coding (CABAC) has been a normative scheme of entropy coding in video coding standards such as H.264, AVS, for its excellent coding efficiency.H.264/AVC and AVS are the new generation of video coding standard, and its compression performance has led it to the most efficient video coding project. So we first carried on the research of H.264/AVC, and put emphasis on its arithmetic coding. By careful analysis, we found that the binarization of the syntax element is a neglected process in CABAC, and the binarization's property will directly affect the coding efficiency. The Unary code is applied in the binarization of the translation coefficients in the national standard AVS. It is preponderant for the small values, but will bring more redundancy for the bigger ones. So the Exp-Golomb code is introduced for the redundancy reduction. This paper presents a algorithm that combined Unary code and Exp-Golomb code. It switched separately in consideration of the difference between luma and chroma coefficients, instead of setting a marking bit which will bing loss at the same time, and also added some corresponding context models. It had been tested on SD, HD and CIF, three types of sequence, and the bitrate had been saved by 1.24134%, 0.10016% and 0.16029% on average. So it is proved that the performance of CABAC could be improved as expected, by integrating the offered binarization algorithm into it.Next, it is observed that it is in a high probability that the abslevel turns out to be 1. So there's a great correlation among the second bitplane. So we offered a new arithmetic coding scheme in which the design of context models is based on level and bitplane. Considering the difference between luma and chroma coefficients, we will make it separately. The 2th order modeling idea of AVS standard is adopted in this scheme, and the context models are mainly partitioned by the bitplane. We will make the context models more detailed for the first and...