IEEE Trans Image Process. 2001;10(3):470-5. doi: 10.1109/83.908536.
A new vector quantizer codebook design for video compression is described. This uses the notion that symmetries in the data, which are seldom captured exactly in any training dataset, are important perceptually and lead to a more robust codebook. The method is illustrated using three-dimensional (3-D) wavelet transformed video sequences.
描述了一种用于视频压缩的新矢量量化码本设计。该方法利用了数据中的对称性概念,这些对称性在任何训练数据集中都很少被准确捕捉,但在感知上很重要,并且可以导致更鲁棒的码本。该方法使用三维(3-D)小波变换视频序列进行了说明。