论文部分内容阅读
在本文中，我们介绍了一种基于离散余弦变换（ＤＣＴ）子空间失真测度的快速矢量编码算法．该算法利用ＤＣＴ子空间映射，将失真测度维数从１６降至４，从而使编码计算复杂度降为１／４，并且结合部分失真算法进一步减小了编码计算复杂度．同时本文提出的最小覆盖集搜索算法大大改善了子空间失真测度算法等非最优矢量编码算法的信噪比特性．测试结果表明，相对于采用欧氏测度的穷尽搜索编码算法，该算法运算时间减少８５—８８％，算术运算数目减少９０％以上，信噪比损失仅为０．０４ｄＢ．
In this paper, we introduce a fast vector coding algorithm based on Discrete Cosine Transform (DCT) subspace distortion measure. The algorithm uses DCT subspace mapping to reduce the dimension of distortion measure from 16 to 4, so that the coding complexity is reduced to 1/4, and the coding complexity is further reduced by combining partial distortion algorithm. At the same time, the minimum cover set search algorithm proposed in this paper has greatly improved the signal-noise ratio characteristics of non-optimal vector coding algorithms such as subspace distortion measure algorithm. The test results show that compared with the exhaustive search coding algorithm using Euclidean metric, the algorithm reduces the computation time by 85-88%, the number of arithmetic operations by more than 90%, and the SNR loss by only 0.04dB.