DCT域网格编码矢量量化

Trellis Coded Vector Quantization in DCT domain

本文提出了在离散余弦变换（ＤＣＴ）域作网格编码矢量量化（ＴＣＶＱ），并利用量化后相关性的新方法。该方法结合了ＤＣＴ变换、矢量验化和网格编码量化的优点，它充分利用信号ＤＣＴ域和时域的相关性来逼近率失真下界，并利用维特比算法来寻找最佳运化序列。同时，它还利用量化后码定间存在的相关性来进一步降低编码年。仿真结果表明，该方法在保持高压缩比的同时具有很高的信噪比性能。在相同编码率下它比全搜索矢量均化（ＶＱ）好ｌｄＢ左右。同时该方法具有计算量适中，解码简单以及对误差扩散不敏感的优点。

A new method which performs Trellis coded vector quantization (TCVQ) in DCT domain is firstly proposed. This method combines the virtues of DCT transformation, VQ and TCQ, and makes use of the correlations in DCT domain and time domain to approach low limit of rate distortion function. Viterbi algorithm is used to fipd a optimized quantization sequence, and the correlation between quantized signatls is utilized to decrease the code rate. Simulation results state it has good quantization performance in SNR While keeping high compression ratio. With the same rate , it is superior to full-search VQ with 1 .OdB or so. At the same time, the method has such capability that it has only modest encoding complexity with simple decoding and its insensitive to channel errors.