基于图非负矩阵分解的图像配准.pdfVIP

第38卷第 12期 光电工程 Vo1．38，No．12 2011年 12月 Opto-ElectronicEngineering Dec，2011 文章编号：1003—501X(2011)12—0137—08 ImageRegistrationBasedonNon—negative M atrixFactorizationofGraoh LENGCheng—cai“，XUW ei，YANW ei-dong“，HELi (a．DepartmentofAppliedMathematics；b．SchoolofAutomation， NorthwesternPolytechnicalUniversity，Xian710072，China) Abstract：A novelimageregistrationmethod isproposed basedon non—negativemarxfactorizationbyusing the absolute~amrevectorsofthespectrum ofgraph．First，thenon—negativewei曲tmatricesoftheundirectedweighted graphbythefeaturepointsofhteimagesareconstructed，from whichthefeaturebasisimageobtainedcancontainthe globalfeaturesoftheoriginalimagesusingnon—negativemarxfactorization．Theinitialvalueofnon—negativemarx factorization isevaluated iteratively using theabsolute featurevectorsofthe spectrum ofhtenon—negativeweight matrices，whichcai1reflectthesturcturalcharacteristicsinformationofthergaph，andimprovetheimagematchingrate． Finally,htecorrectfeaturecorrespondencesarefoundinthefeaturebasisvectorsspacebetweenhteweightedgraphs．The experimentsonboth syntheticimagesandreal—word imagesindicatethattheproposedmethod isfeasibleforimage registration． Keywords：imageregistration；spectralgraph；rgaphmatching；absolutefeautrevectorsofhtespectra；non—negative matrixfactorization(NMF) CLCnumber：TP391 Documentcode：A doi：10．3969／j．issn．1003—501X．2011．12．026 基于图非负矩阵分解的图像配准 冷成财 ，徐 伟a，延伟东a，何 力b (西北工业大学 a．应用数学系；b．自动化学院，西安 710072) 摘要：本文提出一种新的利用图的谱对应绝对值特征向量的非负矩阵分解图像配准方法。首先利用图像特征构造 了无向权图的非负权矩阵，通过非负矩阵分解得到了包含原始图像全部特征的特征基图像；然后将非负权矩阵谱