Optimized Context Quantization for I-Ary Source

Abstract

In this paper, the optimal Context quantization for the source is present. By considering correlations among values of source symbols, these conditional probability distributions are sorted by values of conditions firstly. Then the dynamic programming is used to implement the Context quantization. The description length of the Context model is used as the judgment parameter. Based on the criterion that the neighbourhood conditional probability distributions could be merged, our algorithm finds the optimal structure with minimum description length and the optimal Context quantization results could be achieved. The experiment results indicate that the proposed algorithm could achieve the similar result with other adaptive Context quantization algorithms with reasonable computational complexity.