基于小波变换的最小失真预测/多级矢量量化

矢量量化器的压缩性能随维数的增大而提高，但复杂度亦随维数的增大呈指数增大，限制了大维数矢量的使用。本文利用小波变换产生的子带间的相关性，提出一种新的最小失真预测／多级矢量量化算法。一方面通过最小失真预测来降低时间复杂度，使得编码６３Ｄ的矢量只需付出相当于１５Ｄ矢量的时间复杂度代价；另一方面通过增强多级矢量量化算法来进一步降低复杂度。在复杂度得到极大降低的同时，仍具有很好的编码性能。

Minimun distortion prediction/multistage vector quantization based on wavelet transform

Performance of vector quantizer improves as increase of dimension of vector,but its complexity increases in exponent too,which prohibit use of larger dimension vector In this paper a new algorithm of the minimun distortion prediction/multistage VQ is presented,which on the one hand reduce computational complexity by the minimun distortion prediction that coding 63D vectors only cost expenditures of 15D's,on the other hand diminish complexity by a multistage VQ While complexity is considerably minimized,there is still a better coding performance