小波变换具有良好的空间-频率局部化性能,主要表现在频率压缩特性、空间压缩特性、系数分布的相似性3个方面,这些特性都有利于进行图象压缩.但是早期的小波压缩算法大多没有利用系数分布的相似性.该文借鉴了零树算法和Rinaldo块预测的思想,提出了一种新的旨在压缩重要小波系数结构性冗余的静止图象压缩方法,实验结果证明了这种方法的有效性.

Wavelet transform has very good spatial-frequencial localization characteristics which show itself mainly at three aspects: frequency compression feature, space compression feature and structural similarity of wavelet coefficients among different scales. All these characteristics are propitious to compress still images. But classical compression methods based on wavelet transform seldom make good use of the structural similarity of significant wavelet coefficients. This paper uses the idea of zero-tree compression algorithm and Rinaldo's block predicting method for reference and presents a new still image compression method to eliminate the structural redundancy of significant wavelet coefficients on different scales. The experiment results are satisfactory and prove the validity of this method.