快速码字搜索算法中一维特征量的最佳选择方法

Criterion for Optional 1-D Characteristics Selection in Fast Codeword Search Algorithms of Vector Quantization

矢量量化编码过程中的最近邻码字搜索需要进行大量的矢量间距离的计算,这个过程的计算复杂度极高,严重限制了其实际使用.为了加速矢量量化的编码过程,许多文献提出了各种不同组合的基于均值、2-范数、方差和角度的矢量一维特征量的快速最近邻矢量量化码字搜索算法.通过实验给出了这四个一维特征量单独使用以及相互组合的所有情况下各算法的搜索范围和编码时间,并对它们进行了比较和分析,进而提出了在实际进行编码时如何最优地进行一维特征量选取的准则.

The nearest neighbor codeword search in the encoding process of vector quantization（VQ） needs a great deal of distance computations between vectors,which prevents its practical applications.In order to speed up VQ encoding process,various fast codeword search algorithms have been developed for vector quantization based on 1-D characteristics.These 1-D characteristics of a vector include the average or the mean,the variance,the L2 norm and the angle.Comparisons and analysis are conducted for these fast algorithms by using the four characteristics individually and together in terms of the search space and encoding time.The criterion about how to combine these 1-D characteristics optimally is also given in this paper.