图像匹配结合SIFT和区域匹配程序2012国际会议上控制工.pdf
上传人:qw****27 上传时间:2024-09-12 格式:PDF 页数:3 大小:1.6MB 金币:15 举报 版权申诉
预览加载中,请您耐心等待几秒...

图像匹配结合SIFT和区域匹配程序2012国际会议上控制工.pdf

图像匹配结合SIFT和区域匹配程序2012国际会议上控制工.pdf

预览

在线预览结束,喜欢就下载吧,查找使用更方便

15 金币

下载此文档

如果您无法下载资料,请参考说明:

1、部分资料下载需要金币,请确保您的账户上有足够的金币

2、已购买过的文档,再次下载不重复扣费

3、资料包下载后请先用软件解压,在使用对应软件打开

2012InternationalConferenceonControlEngineeringandCommunicationTechnologyImagematchingcombineSIFTwithregionalSSDAQiuWentao1,2,ZhaoJian1*,LiuJie1(1ChangchunInstituteofOptics,FineMechanicsandPhysics,ChineseAcademyofSciences,Changchun130033,China,E-mail:zhaojian6789@yahoo.com;2GraduateUniversityoftheChineseAcademyofSciences,Beijing100039,China)Abstract—Imagematchingisatthebaseofmanycomputervisionproblems,suchasobjectrecognitionorimagestitching.StandardSIFTprovidespoorperformancewhenimagesunderviewpointchangeconditionsandwithsimilarcorners.Hence,weproposeamatchingalgorithmcombineregionalSSDAwithsimplifiedSIFTalgorithm.WedemonstratethroughexperimentsthatouralgorithmyieldsbetterperformanceinimagesofFigure1SIFTcomputationalprocessviewpointchangeandsimilarfeaturepoints.Besides,thesimplifiedalgorithmcutdownabouthalfthetimewasoriginallyScale-space[7]extremadetection,extremumdetectioninneededinourtestedimages.bothscalespaceandimagecoordinateplane.AGaussianKeywords-Imagematching˗SIFT˗SSDApyramidisconstructedandcandidatepointsareextractedbyscanninglocalextremuminaseriesofDoGI.INTRODUCTION(DifferenceofGaussian)images.Keypointlocation,candidatepointsarelocalizedtosub-Imagematchingisanessentialpartofmanymodernpixelaccuracythroughfittingthree-dimensionquadraticcomputervisionsystemsthatsolvetaskssuchasobjectorfunction,andunstablepointsoflowcontrastorstrongscenerecognition,stereocorrespondence,imageindexingandedgeresponseareeliminated.itisthebasisofimagestitching,imagefusion[1,2].ItisaOrientationassignment,orientationsareassignedtoeachtechniquethatwildlyusedinprocessingmedicalimages,keypointlocationbasedonstatisticalresultoflocalremotesensingimagesetc.Imageregistrationistofindaimagegradientdirections.transformationbetweentwoortwomoreimages,whichisofKeypointdescriptor,SIFTdescriptoristherepresentationthesamesceneobtainedfromdifferenttimeordifferentofstatisticallocalimagegradientswhichmeasuredatthesensorsordifferentviewpoint,enablestheiridentica