【作者】 张雷；

【导师】 陈向东；

【作者基本信息】 西南交通大学 ， 通信与信息系统， 2006， 硕士

【摘要】 图像是一种重要的二维信号,由于其数据量很大,在存储和传输的时候要对其进行压缩处理。小波变换是一种新兴的数学工具,基于小波的图像压缩技术正受到广泛的关注和研究。图像经过小波变换以后,在时域和频域都具有良好的局部化特性,重建图像中可以克服采用离散余弦变换编码所固有的方块效应,而且与人类视觉特性相一致。 本论文主要研究了基于小波变换的静态图像压缩算法,完成了以下一些工作:介绍了传统图像编码和小波变换编码的原理和方法,列举了常用的图像压缩标准,在理论分析的基础上总结出小波变换的优势。重点研究了小波编码算法中著名的嵌入式零树小波算法(EZW)及其改进形式SPIHT算法,并在其基础上其提出了自己的改进算法,包括采用提升方案,降低了原始小波变换的复杂度;在量化时引入了符合人类视觉特性的Mannos模板重新定义各子带系数幅值以提高重构图像的主观质量;为了进一步降低复杂度,在编码时对EZW算法采用新的阈值循环策略,而在进行SPIHT编码时将低频子带的系数全默认为正,减少了输出的码字,提高了编码效率。 试验结果表明,虽然本文所提出的改进算法与经典的EZW和SPIHT相比虽然在峰值信噪比上略有下降,但是算法的复杂度有所降低且使编码的方式更贴近人类视觉系统,总体来说是一种有效的压缩算法。更多还原

【Abstract】 Image is an important two-dimension signal. Because of the huge data it contains, image must be compressed when it is stored or transported. The wavelet transform (WT) is a rising mathematical tool. The technology of image compression based on wavelet has drawn much attention and has been researched broadly. When a image is transformed by the wavelet, it has favorable localize characteristic in both time-domain and frequence-region. And in the rebuild image the connatural diamonds affection of the discrete cosine transform can be overcame. Furthermore, the WT is identical to Human Visual System(HVS).This paper is mainly about the static image compress algorithm based on wavelet transform, and complete these work: Introduced the theory and technique of traditional image coding and WT coding. Enumerated the prevalent image compressional standards. This paper focuses on the famous WT coding algorithm which named embedded zerotree wavelet (EZW) algorithm and its improved form SPIHT algorithm, based on EZW and SPIHT, an improved algorithm is introduced which includes using the lifting scheme to reduce the complexity of the original WT. To improve the subjective quality, use the Mannos template which accord to HVS to reset the magnitude of the coefficients for each sub-image. To reduce the complexity more, use new threshold scheme in EZW process, default the sign of coefficients in low frequence subimage to positive in SPIHT process, so the deferent code cut down and the coding efficiency improved.The simulation results show that, although the improved algorithm has a lower performance on PSNR compared with classical EZW and SPIHT, its complexity decreases and the way of its coding is more in accord with Human Visual System, so it is more effective.更多还原