基于零树分类的小波图像二维网格编码量化

提出了一种小波图像在进行小波系数零树分类后运用二维网格编码量化(2D-TCQ)的新方法。首先根据小波图像中各子带系数固有的树结构关系对其进行零树分类,然后在扩展的二维码书空间对重要性系数运用网格编码量化,利用卷积编码和信号空间扩展进一步增大量化信号间的欧氏距离,并用维特比算法寻找最优量化序列。仿真结果表明,该方法比小波零树分类后相同码书尺寸下使用 TCQ信噪比获得了0.6dB左右的改善。使用小一倍的码书时,2D-TCQ比TCQ获得了0.1dB左右的改善,而且降低了编码率。由于本方法可以采用较小的码书尺寸,所以可以应用到低存储、低功耗的编解码环境。同时该方法具有编码计算复杂度适中、解码简单的优点。

2D-TCQ of Wavelet Coefficients Using Zero-Tree Classification

A new method of two-dimensional trellis coded quantization(2D-TCQ) of wavelet coefficients based on zero-tree classification is proposed. First,the zero-tree classification is made by use of the correlation between the subbands,then,a small codebook is expanded to a larger virtual codebook and Viterbi algorithm is adopted in two dimensions to search for the optimum quantization order. Simulation results show that with this method the SNR is improved by 0.6dB or so over TCQ after zero-tree classification in the same codebook size. When 2D-TCQ has a codebook size as half as TCQ,an improvement of 0.1dB or so over TCQ is achieved after zero-tree classification,and the coding rate is reduced. By using a small-size codebook,2D-TCQ can be applied in many low-power encoding/decoding conditions. The method also has such advantages as modest encoding complexity and simple decoding.