基于变步长分组的H.264系数码表优化

H.264标准采用内容自适应变长编码CAVLC,针对不同的编码系数采用多组码表分别进行编码,在提高熵编码效率的同时,结构各异的码表使解码算法的复杂度过高.本文通过分析码表的结构特点和码字查找的统计规律,提出了一种基于变步长分组的系数码表优化方法,将遍历二叉树的全搜索过程简化为分级逐层查找.实验结果表明,本文给出的快速CAVLC解码算法与原有解码算法相比,可以提高解码速度达3倍以上,且编码码率越高,算法的优化性能越好.

Optimization of the H.264 coefficient tables based on variable step-length group

Context-adaptive variable length coding (CAVLC) is chosen by H.264 video coding standard as entropy codec. To the difference coefficients, CAVLC defines many tables and uses the different tables dynamically for higher coding efficiency, but complicated code-tables result in higher computation. So a high performance-decoding algorithm using a new variable step-length group approach is proposed for H.264 coefficient tables in this paper. Through analyzing the structure of code-tables and statistic characteristic of code matching, the similar prefix codes are classified into one group to avoid the full-tree search process. The computer simulation shows that the proposed algorithm can enhance the whole decoding speed 3 times faster than full search algorithm, and if the bitrate is higher, the result is better.