CAJ | 학술논문

使用聚类算法实现 Context 量化不仅可以推广量化器的应用范围，而且可以获得编码性能较理想的优化量化器。然而，聚类算法依赖于相似测度。前期研究中采用的描述长度增量不能完全满足相似测度的各项属性，从而导致聚类结果的性能偏差。因此，提出数学描述特性更好的奇异测度增量作为两个计数向量的相似测度，并说明其相应性质。实验结果证明，使用奇异测度增量作为相似测度，不仅能够保证 Context 量化器的稳定性，而且还获得更佳的编码结果。
사용취류산법실현 Context 양화불부가이추엄양화기적응용범위，이차가이획득편마성능교이상적우화양화기。연이，취류산법의뢰우상사측도。전기연구중채용적묘술장도증량불능완전만족상사측도적각항속성，종이도치취류결과적성능편차。인차，제출수학묘술특성경호적기이측도증량작위량개계수향량적상사측도，병설명기상응성질。실험결과증명，사용기이측도증량작위상사측도，불부능구보증 Context 양화기적은정성，이차환획득경가적편마결과。
The context quantization based on the clustering algorithm can not only improve the application range but also get the optimizing quantizer with ideal coding efficiency.However,the clustering algorithm relies on the similarity measure.In the previous research,the increment of the description length is not suitable for the various attributes of similarity measure so as to cause the deviation of cluster result.So the increment of amazing measure with better mathematic descriptive feature as the similarity measure of two count vector quantity is proposed and its corresponding quality is stated.The results indicate that the application of increment of amazing measure as the similarity measure can not only guarantee the stability of Context quantizer,but also achieve better coding efficiency.