相关性度量及其在超谱图像处理中的应用博士论文.docVIP

摘　　要 图像融合及图像压缩是遥感超谱图像处理中用到的重要技术。图像融合技术为利用超谱数据的高数据维进行特征提取和分类提供了新的途径。图像融合技术的快速发展，要求用系统的性能评估方案来客观地评价图像融合的质量、指导融合、提高融合效果。图像压缩是解决超谱图像的大数据量给数据传输和保存所带来压力的有效方法。本文以遥感超谱图像的处理为背景，研究图像的融合、压缩及性能评估方法。从信息相关性的角度出发对不同图像处理方法的性能进行评估是一个新途径，本文对其关键技术多变量非线性相关性的定量度量理论进行了研究。本文的主要内容包括： 1）提出用于定量描述变量之间的非线性相关性的非线性相关系数和非线性相关信息熵的概念。其中，非线性相关系数不仅能够两个变量之间的普遍关系，而且能够以闭区间[0，1]的值来表征该相关程度的强弱。并且1表示两个变量之间最强的相关程度，0表示两个变量之间没有统计意义上的相互关系。相关信息熵以多变量系统的整体相关性作为研究对象，描述的是一个系统中各个变量间相互关联和信息重叠的程度，物理意义明确；以[0，1]之间的数量化系统信息相关的程度，易于进行比较和分析，描述更加直观；以经典相关分析中相关系数的概念和信息论中信息熵函数为基础，计算简单速度快Image fusion technology provides a new approach for feature extraction and classification of hyperspectral data utilizing the high dimensionality of hyperspectral data. The fast development of image fusion technology requires that the objective performances evaluation method should be used to assess the fused image quality and indicate the fusion process and improve the fusion results. Image compression is an effective technique to resolve the transmission and storage problem caused by the large data volume of hyperspectral image. This dissertation focuses on the image fusion, image compression and their performances evaluation techniques, which will be used in the hyperspectral data processing. To assess the performances of different image processing techniques from the information correlation point of view is a new path, its key technique, i.e. the measurement of nonlinear correlation among a multi variable data set is studied. The main contents of this dissertation are as follows: The nonlinear correlation coefficient and nonlinear correlation information entropy concepts are proposed to quantitatively describe the nonlinear correlation among variables. The nonlinear correlation coefficient not only can describe the general relationship between two variables as the mutual information does, but also can use the value in a closed interval [0, 1] to indicate the degree of such relationship. Moreover, value 1 indicates the strongest relationship b