| 引用本文: | 陈浩,宿腾野,滑艺,刘东.SLWE概率估计方法在区间编码中的应用研究[J].哈尔滨工业大学学报,2016,48(5):43.DOI:10.11918/j.issn.0367-6234.2016.05.006 |
| CHEN Hao,SU Tengye,HUA Yi,LIU Dong.The application for probability estimation of SLWE on range coder[J].Journal of Harbin Institute of Technology,2016,48(5):43.DOI:10.11918/j.issn.0367-6234.2016.05.006 |
|
| 摘要: |
| SLWE概率估计方法具有较强的适应非平稳数据能力，为拓展其在熵编码中的应用，更有效地编码非平稳数据，设计在区间编码上的应用方案.首先针对概率估计模型替换时SLWE估计出的概率如何映射到区间的问题，不进行概率更新的计算，而是基于SLWE思想直接更新各字符所占区间大小，再根据区间编码中总区间上下界计算方法调整总区间.既结合SLWE应对非平稳数据的优势，又避免概率运算.同时，针对更新各字符所占的整型数据区间后字符所占区间大小可能小于1导致编码字符丢失的问题，采用设定每种字符最小区间作为阈值的控制方法.对非平稳数据编码的实验结果表明，基于SLWE的区间编码比基于加窗法等传统概率估计方法的压缩率要高出1%~5%.
|
| 关键词: 熵编码 非平稳数据 随机学习弱估计 概率估计 区间编码 |
| DOI：10.11918/j.issn.0367-6234.2016.05.006 |
| 分类号:TN911.73 |
| 文献标识码:A |
| 基金项目:国家863项目2012AA12A405;国家自然科学基金61102159. |
|
| The application for probability estimation of SLWE on range coder |
|
CHEN Hao, SU Tengye, HUA Yi, LIU Dong
|
|
（School of Electronics and Information Engineering, Harbin Institute of Technology, 150001 Harbin, China）
|
| Abstract: |
| The probability estimate method of SLWE can adapt to the non-stationary data. In order to expand its application in entropy coding and code non-stationary data more effectively, an application scheme of SLWE in the range coding is designed. Firstly, to tackle the problem how to map the estimated probability by SLWE to the coding interval, instead of calculating the update probability, we propose to update the range of every character directly based on the idea of SLWE and then adjust the total range according to the computing method of the upper and lower range bounds for range encoding. It not noly combines the advantage of SLWE to cope with the non-stationary data, but also avoids the probability calculation. In addition, the coding range after the update for every character may be less than 1, which causes the loss of the character. To solve this problem, we present a control method to set the minimum range of each character. Experimental results for non-stationary data coding show that the SLWE-based range coder achieves 1%~5% higher than that using traditional probability estimation (e. g. windowing method) in terms of compression ratio.
|
| Key words: entropy coding non-stationary data stochastic learning weak estimator probability estimate range coder |
