基于描述长度的Context建模算法

Context Modeling Based on Description Length

在基于Context建模的熵编码系统中,为了达到预期的压缩性能,需要通过Context量化来缓解由高阶Context模型所引入的"Context稀释"问题。为此,该文提出一种通过最小化描述长度来实现Context量化(Minimum Description Length Context Quantization,MDLCQ)的算法。该算法使用描述长度作为评价准则,通过动态规划算法来实现单条件的最优Context量化,然后通过循环迭代来实现多条件的Context量化。该算法不仅可以得到多值信源的优化Context量化器,而且可以自适应地确定各个条件的重要性从而确定模型的最佳阶数。实验结果表明:由MDLCQ算法所得到的Context量化器,可以明显改善熵编码系统的压缩性能。

In entropy coding systems based on the context modeling, the ＂context dilution＂ problem introduced by high-order context models needs to be alleviated by the context quantization to achieve the desired compression gain. Therefore, an algorithm is proposed to implement the Context Quantization by the Minimizing Description Length（MDLCQ） in this paper. With the description length as the evaluation criterion, the Context Quantization Of Single-Condition（CQOSC） is attained by the dynamic programming algorithm. Then the context quantizer of multi-conditions can be designed by the iterated application of CQOSC. This algorithm can not only design the optimized context quantizer for multi-valued sources, but also determine adaptively the importance of every condition so as to design the best order of the model. The experimental results show that the context quantizer designed by the MDLCQ algorithm can apparently improve the compression performance of the entropy coding system.