【作者】 黄向生；

【导师】 杨小帆；

【作者基本信息】 重庆大学 ， 计算机软件与理论， 2002， 硕士

【摘要】 图像压缩在多媒体存储和传输中起着非常重要的作用，图像压缩的研究成为理论界和工程界广泛关注的课题，其中基于小波变换的图像压缩已经成为图像压缩研究的主流。由于小波本身正处在一个不断发展的阶段，基于小波变换的图像压缩也处于探索之中。本文首先介绍小波分析及其性质，对小波母函数、尺度函数、多分辨分析等进行分析。然后引入最新的小波变换，阐述第二代小波的原理和构造方案--提升方案，并且简单叙述在第二代小波的理论框架下的整数到整数小波变换和形态小波变换。由于整数到整数小波变换具有变换不失真和变换速度快等优点，因此，本文的图像压缩研究基本上是采用整数小波变换。本文就以下两个方面对基于小波变换的图像压缩进行研究：小波基的选取。由于不同的图像具有不同的特点，如何选取一个小波基使其与图像相匹配，从而达到较好的压缩效果，是一个很值得研究问题。本文在S+P变换的基础上，根据图像的熵提出一种自适应可逆整数小波变换方案。实验结果表明，该算法比典型S+P变换的压缩比高。小波变换系数的编码方案。本文对EZW、SPIHT和EBCOT等三种比较典型的小波变换系数编码进行分析。EZW和SPIHT这两个算法充分利用小波变换系数的不同子带之间的相关性，构造四叉树进行变形的变长编码。而EBCOT是利用同一子带内相邻系数的相关性，构造一种良好的上下文模型进行高阶自适应算术编码。在分析这些编码的基础上，本文提出一种基于偏序零树Markov显著性的小波变换系数编码方案，充分利用小波变换后的系数所具有的子带之间的偏序性、自相似性和同一子带内的相邻系数的Markov显著性的特点。实验表明，该编码方案在大多数情况下优于前面三种典型的编码方案。更多还原

【Abstract】 Image compression technique plays an important role in efficient transmission and storage of multimedia data. Therefore, academic and engineering circles have been paying much attention to the research and development of image compression for many years. Because of its multiresolution properties, wavelet transform has been successfully applied as the decorrelation transform in image compression. Since wavelet is still in its childhood, however, its power in the image compression application field is far from been fully expoited.The thesis introduces the basic concepts of wavelet transform and multiresolution analysis. Then the second generation wavelets and its construction scheme ?lifting scheme ?are presented. With the help of lifting scheme, wavelet transforms that map integers to integers and morphological wavelet are built up.The main works and conclusions in this thesis are listed as follows:? The selection of wavelet basis.Because different images have different characteristic properties, it is worthy to do research on how to select wavelet basis for the compression of a specific image. A novel image compression algorithm, which adaptively takes reversible integer-to-integer wavelets as the decorrelation transform depending on the value of entropy of image, is presented. The experimental results show that the algorithm can achieve higher lossless compression ratio than the classical S+P transform in most cases.? The modeling of the wavelet coefficients.Three typical coding schemes based on wavelet transform are analyzed. Both EZW and SPIHT utilize the correlation of different sub-band to construct a quad-tree, which is a variation of the classical run-length-coding, while EBCOT makes use of the correlation of neighbors in the same sub-band to organize context for high order adaptive arithmetic coding. Finally, we propose an algorithm for image compression based on the partial ordering zero-tree and the approximate Markov significance of neighbors, which makes full use of the properties of the coefficients of wavelet transformation. The experimental results demonstrate that in the case of lossless compression, the algorithm achieves higher compression ratio than the three coding schemes.更多还原