一种基于整数小波变换的图像编码算法  

An Image Compression Algorithm Based on Integer Wavelet Transform

作　　者：张立保[1] 王珂[1] 

机构地区：[1]吉林大学通信工程学院

出　　处：《软件学报》2003年第8期1433-1438,共6页Journal of Software

基　　金：国家自然科学基金~~

摘　　要：针对静止图像经整数小波变换(integer wavelet transform)后,各子带系数的幅值动态变化范围小,不利于零树编码(embedded zerotree wavelet coding)的缺点,采用“从1开始的整数平方”代替“2的整数次幂”作为零树编码的量化阈值,缩短了量化阈值间隔,通过减少重要系数在较低量化阈值中出现的机会,增加了编码过程中的零树数量;同时通过减少参与编码的字符数及对最高频带零树不编码,简化了编码过程。实验结果表明,整数平方量化阈值下的零树编码(integer square zerotree wavelet coding)解决了整数小波变换中零树编码的低效问题,提高了静止图像的编码效率。Integer wavelet transform (IWT) and IWT-Based image-compression offer many advantages over traditional wavelet. But every subband coefficients of the IWT-Image have smaller dynamic change value and worse energy compaction than the first generation wavelet. Those reduce efficiency of embedded zerotree wavelet (EZW) coding. This paper proposes 'integer square quantization threshold from 1' instead of '2 of integer power quantization threshold' during EZW coding. It shortens the distances of every quantization threshold, reduces probability of the important coefficients in the lower quantization threshold area and increases zerotrees. It reduces the coding signs and does not code the zerotrees of the most high-frequency subbands. The experiments results show that integer square zerotree wavelet (ISZW) coding algorithm solves the low efficiency of EZW based on IWT and improves the performance of compression algorithm. It may be used in still picture compression and transmission.

关 键 词：整数小波变换 零树编码 整数平方零树编码 2的整数次幂 量化阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]