图像融合中的特征选取及分类与决策  

Feature Selection, Classification and Decision in Image Fusion

作　　者：冉向书[1] 任金昌[1] 

机构地区：[1]西北工业大学计算机学院,陕西西安710072

出　　处：《计算机应用研究》2005年第9期170-172,共3页Application Research of Computers

基　　金：航空基金资助项目(01F53036)

摘　　要：图像融合在遥感等许多领域具有重要作用。简要分析了预处理和像素级别融合的技术,着重探讨了用于特征级和决策级融合的关键技术,包括特征提取与选择、分类器设计及决策等。针对几种典型的有监督及无监督分类算法,进行了深入的分析和比较。最后指出了在分类决策方面图像融合技术发展的方向。Image fusion has important applications in many military and civil fields such as remote sensing. In this paper, preprocessing and pixel-level fusion techniques are firstly analyzed, and then fusion techniques of feature-level and decisionlevel are emphasized, including feature extraction and selection, classifier design along with decision. Comprehensive analysis and comparisons are given for several typical algorithms in supervised and unsupervised classification. Finally, future directions are prospected in classification and decision of image fusion.

关 键 词：图像融合 特征选取 分类与决策 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]