PAIGE = PAirwise Image Geometry Encoding for Improved Efficiency in Structure-From-Motion (CVPR 2015)

PAIGE = PAirwise Image Geometry Encoding for Improved Efficiency in Structure-From-Motion (CVPR 2015)

ID:40363353

大小:3.00 MB

页数:10页

时间:2019-08-01

PAIGE = PAirwise Image Geometry Encoding for Improved Efficiency in Structure-From-Motion (CVPR 2015)_第1页
PAIGE = PAirwise Image Geometry Encoding for Improved Efficiency in Structure-From-Motion (CVPR 2015)_第2页
PAIGE = PAirwise Image Geometry Encoding for Improved Efficiency in Structure-From-Motion (CVPR 2015)_第3页
PAIGE = PAirwise Image Geometry Encoding for Improved Efficiency in Structure-From-Motion (CVPR 2015)_第4页
PAIGE = PAirwise Image Geometry Encoding for Improved Efficiency in Structure-From-Motion (CVPR 2015)_第5页
资源描述:

《PAIGE = PAirwise Image Geometry Encoding for Improved Efficiency in Structure-From-Motion (CVPR 2015)》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库

1、PAIGE:PAirwiseImageGeometryEncodingforImprovedEfficiencyinStructure-from-MotionJohannesL.Schonberger,AlexanderC.Berg,Jan-MichaelFrahm¨DepartmentofComputerScience,TheUniversityofNorthCarolinaatChapelHillfjsch,aberg,jmfg@cs.unc.eduAbstractconnectionsacrosstheentirescene.Howeve

2、r,exhaus-tivelysearchingfortheseoverlappingpairsisinfeasibleLarge-scaleStructure-from-Motionsystemstypicallyforlarge-scaleimagecollectionsduetoquadraticcompu-spendmajorcomputationaleffortonpairwiseimagematch-tationalcomplexityinthenumberofimagesandfeatures.ingandgeometricve

3、rificationinordertodiscovercon-Moreover,asthenumberofregisteredimagesgrows,thenectedcomponentsinlarge-scale,unorderedimagecollec-scalabilityofbundle-adjustmentalgorithmsbecomesasig-tions.Inrecentyears,theresearchcommunityhasspentnificantperformancebottleneck.significantefforto

4、nimprovingtheefficiencyofthisstage.ThispaperevaluatesexistingtechniquesforreducingtheInthispaper,wepresentacomprehensiveoverviewofcostofStages2and3,featurematchingandgeometricvariousstate-of-the-artmethods,evaluatingandanalyzingverification.Usually,themajorityofimagepairsinun

5、-theirperformance.Basedontheinsightsofthisevalua-orderedInternetphoto-collectionsdonothavesceneover-tion,weproposealearning-basedapproach,thePAirwiselap,sorejectingthosepairsdominatesexecutiontime,evenImageGeometryEncoding(PAIGE),toefficientlyidentifythoughsuchpairsarenotuse

6、fulfor3Dreconstruction.Con-imagepairswithsceneoverlapwithouttheneedtoper-sequently,variousapproacheshavebeenproposedtoeffi-formexhaustiveputativematchingandgeometricverifica-cientlyfindoverlappingpairsinnoisydatasetsandonlyfor-tion.PAIGEachievesstate-of-the-artperformanceandin

7、-wardthosepairstoStages2and3.AdownsideofsendingtegrateswellintoexistingStructure-from-Motionpipelines.fewerimagepairstoStages2and3isthatenoughimageswithoverlappinggeometrymustbeprocessedtoproduceaccuratecameraalignmentandcompletereconstructions.Hence,itisessentialtofindtheri

8、ghttrade-offbetweencom-1.Introductionputationalefficiencyandsufficientimageconnectiv

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。