English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

When using the clustering algorithm, the implementation of the context quantization not only expands the application range of the quantizer, but also obtains better coding performance. However, those clustering algorithms depend on the choice of the similarity measure. In the the previous works, the increment of the description length was suggested but the result is that it cannot fully meets the similarity measure of various attributes, resulting in the performance of the clustering result deviation. In this paper, a new similarity measure which holds better mathematical description is given. The increment of the amazing measure, which denotes the similarity measure, two count vectors are discussed in this paper and its corresponding properties are also explained. The experimental results indicate that when using the proposed similarity measure, both the stability of the context quantizer and the corresponding coding results can be optimized at the same time. Introduction Efficient context modeling entropy coding technique [1] always plays an important role in the field of digital signal compression. Modeling is the key to enhancing the compression efficiency, which takes advantage of the correlation between source symbols as much as possible. By conditioning for the current symbol, the entropy of the conditional probability distribution is reduced. Actually, in practice, these distributions are estimated through the counting approach and then used to drive the arithmetic encoder for assigning the code word for each symbol. Theoretically, the entropy of the conditional probability distribution is the lower band of the code length for coding every symbol. It means that if the entropy is low, coding compression effect is more ideal. However, there are a variety of factors influencing the coding performance. One important of them is the context dilution. In training or coding process, each count vector obtains its corresponding data and the number of these data is used to estimate distributions [2]. If the number of data is not enough to achieve the distribution with high efficiency, the distribution will be trend to the uniform distribution which holds the highest entropy. The context quantization is an efficient approach to tackle the context dilution. Actually, the context dilution makes the coding result including high model cost [3]. In order to make each count vector containing more number of data, integrating some count vectors into one is an ideal method. It is also the way what context quantization does. The first context quantizer is CALIC algorithm in [4], which uses the empirical quantizer to partition the space of the context. After modifying, CALIC could achieve ideal coding results. The strict optimized context quantizers are the minimum conditional entropy context quantization (MCECQ) in [5,6] and the minimum adaptive code length context quantization (MCLCQ) in [7,8]. However, these two quantizers have their flaws. MCECQ relies on the input of the quantization levels, which implies that MCECQ cannot achieve the optimized class adaptively. MCLCQ cannot obtain the optimized results for I-ary source only for binary case. Actually, in [9, 10], the context quantization based on the clustering algorithms are suggested. In [9], the modified genetic algorithm is used to cluster the context model for achieving the quantization levels adaptively. In [10], the increment of the description length is defined as the similarity measure between two conditional probability distributions and the affinity propagation is employed to implement the context quantization. Both of these two quantizers can achieve the similar results with the previous 3rd International Conference on Management, Education, Information and Control (MEICI 2015) © 2015. The authors Published by Atlantis Press 1013 optimized context quantizers. However, the increment of the description length is not the strictly similarity measure, which limit the performance of the coding system. Actually, in the derivation of the increment of the description length, there are new items to be considered. In this paper, researchers call it the amazing measure. Meanwhile, the increment of the amazing measure is also discussed in the the present work, including the features which indicate that the increment of the amazing measure is suit for judging whether two conditional probability distributions are similar. Then the increment of the amazing measure is used for the context quantization to improve the coding efficiency. Amazing Measure In the context modeling entropy coding, the conditional probability distributions are unknown. Usually, these distributions are estimated by using their corresponding count vectors. When a count vector } , , , { 1 1 0   I n n n  V is given, the estimator is described by (1)

The Design of the Context Quantizer on the basis of Amazing Measure