基于区域增长的立体匹配算法分析

基于区域增长的立体匹配算法分析

ID:34594214

大小:11.48 MB

页数:55页

时间:2019-03-08

基于区域增长的立体匹配算法分析_第1页
基于区域增长的立体匹配算法分析_第2页
基于区域增长的立体匹配算法分析_第3页
基于区域增长的立体匹配算法分析_第4页
基于区域增长的立体匹配算法分析_第5页
资源描述:

《基于区域增长的立体匹配算法分析》由会员上传分享,免费在线阅读,更多相关内容在行业资料-天天文库

1、万方数据AbstractEyes,isthemainrouteofhumanperceptionfromcircumstances.Thecomputersnotonlyacquire,recognizeandgettheinformationfromthree·dimensionalscenebymodelingthehumanvisionsystems,butalsomakerelatesanalysis.ComputervisionCanworkinavarietyofdangeroussituationsinsteadofthehuman

2、eyes.Soresearchesonmachinevisionisofquitesignificance.StereomatchingisafocusandchMlengeincomputervision,andhowtoachievestereomatchingquicklywithhigh-precisionisthekeypointformanyscholarsinrecentyears.Becausestereomatchingalgorithmbasedonregiongrowinghasmanyadvantages,thispape

3、radoptsittodoappropriatestereomatchingforimages·Mainworkofthispaperareasfollow:1Illuminationvariationhasacertaininfluenceonthepixelgrayofimage.ThematchingerroroftraditionalCensusTransformalgorithmresultsfromtheover-relianceonthecentralpixel’Sgray.Sothispaperputsforwardanimpro

4、vedCensustransfoITnalgorithmwithhighrobust.Firstly,asumofgrayavqrageandlocalcontrastisappliedtoreplacethecenterpixel’Sgrayvalueintransformwindow,whichimprovethenoiseresistanceoftransformationresultsandthedifferencebetweendifferentblocks.Next,itintroducestheGaussiantemplateint

5、otlleCensustransformationtoenhantetheweightofpixelsclosedtocenterpixel,whicheffectivelyreducedtheimpactofdiscontinuousareaonthematching.Finally,itaddsthesimulationlightingtotheimagesofMiddleburyandemploystheactualmatchingvaluewhichisofferedbyMiddleburytocalculatethematchingpr

6、ecisioninthispaper.Experimentalresultsshowthatthealgorithmhasahighermatchingaccuracyandastrongerrobustnessthantraditionalalgorithm.21Thepaperhasstudiedthemechanismandtheexpansionstepsofregiongrowingalgorithm.Confrontedwithlowtextureimagebasedontraditionalregiongrowingmatching

7、algorithm,thegrowingCan’tbeofwiderangebecausetheseedpoints’numberisnotenough.SoamatchingalgorithmbasedonafusionofSIFTfeaturesandcontourinformationisputforward.Firstofall,theSIFTfeaturepointsandtheinformationofthecontourareextracted.Then,itusestwostepsofmatchingmethodtofinisht

8、hejobofmatchingcontourpoints:thefirststep,thefeaturepointsandepipola

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。