零树框架下整数小波图像编码的改进

整数小波变换(Integer Wavelet Transform)有许多优点，但是图象经整数小波变换（IWT)后，能量集中性较第一代小波变换差很多，不利于嵌入式零树编码(Embedded Zerotree Wavelet Encoding)。因此本文提出一种新算法，从两方面加以改进。首先，采用“整数平方量化阈值选取算法”，根据整数小波变换后各子带系数幅值的动态变化较小，小波图像能量较一般小波差的特点，选取从1开始的正整数平方作为量化闽值的同时引入可调节的量化阈值系统，根据图像中不同区域的重要性选取与之相应的量化阈值，从而增加了零树的数量；其次，提出基于索引表和游程编码的小波零树编码的新思路，简化了编码与解码的过程。实验表明，本文算法充分的将整数小波变换与零树编码结合在一起，改善了压缩质量，提高了压缩效率。

An Improved Integer Wavelet Image Coding Based on Zerotree Framework

Integer wavelet transforms have many advantages. However, their energy concentration performance is much worse than that of common wavelet image transform algorithms. This is not beneficial to embedded zerotree wavelet encoding. For this reason, a novel algorithm is proposed to deal with this problem from two aspects. First, because subband coefficients by integer wavelet transform have smaller dynamic change value and worse energy concentration, a new method called ?integer square quantization threshold algorithm ?is adopted, in which quantization thresholds are chosen from the square of integers beginning with 1. At the same time, the quantization threshold is adjustable according to the importance of the interested region in the image. Secondly, a new approach based on index table and RLE coding is proposed to simplify the coding and decoding processes. Experiments demonstrate that the new algorithm improves not only compression quality, but also compression efficiency.