基于四叉树分类的小波图像二维网络编码量化

2D-TCQ of Wavelet Image Using Quadtree Classification

本文提出了一种小波图像在进行小波系数四叉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的带间相关性对其进行四叉树分类,然后在扩展的二维码书空间对重要类系数进行网格编码量化,运用维特比算法寻找最优量化序列,最终形成有序的嵌入式编码比特流。仿真结果表明,该方法在相同编码率下与SPIHT算法相比,PSNR获得了0.4dB左右的改善;使用小一倍的码书,相同编码率下比四叉树分类后使用一维TCQ获得了0.1dB左右的改善。由于本方法可以采用较小的码书尺寸,计算量较小,所以适用于低存贮、低功耗的编解码环境。

A new method which performs two dimension Trellis Coded Quantization (2D-TCQ) of wavelet image based on quadtree classification is proposed. First, the quadtree classification utilizes correlation between the subbands, then, expands the small codebook to larger virtual codebook and uses Viterbi algorithm in two dimension to search the min distortion quantized route. Ordering embedded output bit stream is formed. Simulation states that this method has an advantages of0.4dB or so over SPIHT. If 2D-TCQ has a codebook size as half as TCQ, this method has an advantages of 0.1dB or so over TCQ after quadtree classification. By using small size codebook, 2D-TCQ can be applied in many low-power encode/decode condition.