一种多光谱图像纹理特征提取及比较方法  

A New Method for Texture Feature Retrieval and Comparison of Multi-spectral Images

作　　者：张斌[1] 刘磊[1] 敬忠良[1] 薛阳[1] 肖刚[1] 

机构地区：[1]上海交通大学电子信息与电气工程学院航空航天信息与控制研究所,上海200030

出　　处：《计算机工程与应用》2005年第26期30-33,113,共5页Computer Engineering and Applications

基　　金：国家自然科学基金项目(编号:60375008);国家科技攻关计划重点项目(编号:2004BA908B07);航空科学基金(编号:02D57003);高校博士点基金(编号:20020248029);航天支撑技术基金资助

摘　　要：该文提出一种基于最小二乘及区域分割的多光谱图像纹理特征提取与比较的方法。通过最小二乘法得到系数矩阵来描述多光谱图像的纹理特征,利用区域分割技术将不同纹理特征的图像区分开,并将相似纹理的图像合并为一类。提取区域的统计特性得到纹理特征提取的结果,并采用四条标准来比较两幅多光谱图像特征的差异程度,最后进行算法仿真,取得了较好的效果。A new method based on least squares and region segmentation is proposed to retrieve and compare the texture features of multi-spectral images in this paper.Vectors of coefficients achieved by least squares method can express the texture of multi-spectral images,Then blocks of different texture are divided separately and blocks of the same texture are merged into the same class using the region segmentation method.The statistical features of each region are retrieved to make up of the texture features of the multi-spectral images.Finally,four rules are proposed to show the difference between two multi-spectral images,The result of simulation shows that it is a practical and effective method,

关 键 词：纹理 多光谱图像 最小二乘法 区域分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]