基于小波变换的零森林图像压缩方法

Image Compression Using Zero-forests of Wavelet Coefficients

该文从理论和实验上证明了同级别不同方向子带的零树位置具有较大相关性,据此提出了嵌入的零森林编码方法MRZF。该方法把森林作为系数表示的基本数据结构,使相邻系数之间,同方向子带之间的相关性以及不同方向子带间的相关性能被有效利用,编码效益更高。该方法可直接输出0,1比特流,编解码速度快,且具有多率特性和低计算复杂性的特点,可实现码流的累进传送。采用国际上通用的标准测试图像对提出的图像压缩方法进行测试,实验结果优于现有方法。

This paper proves the correlations between the position of zero-trees in different sub-bands of same level theoretically and experimentally. Using forests in graph theory to represent zero-trees rooted in different oriented sub-bands of same level, a new image compression algorithm has been proposed, which can exploit efficiently not only the correlation between adjacent coefficients, sub-bands of same direction, but also the different oriented sub-bands in same level. The proposed coding and decoding procedures are fast, have lower computational complexity and efficient performance. The generated stream is completely embedded. The method proposed has been tested on popular standard test images at various rates. The experimental results show that the new method MRZF outperforms the known coder.