文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > IT/计算机 > Multi-level Partition of Unity Implicits图形学
侵权投诉

Multi-level Partition of Unity Implicits图形学

multi-level 思想在图形分割的应用的应用

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Figure12:Left:SmoothblendingoftheStanfordbunnyandCyberwareIgeamod-els.Right:offsettingoftheStanforddragonmodel;f=0.0075(top)andf= 0.01(bottom);α=0.

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

75.

Figure14:LinearmorphingoftwoheadmodelsapproximatedbyMPUimplicits,MaxPlanckHeadandHeadofMichelangelo’sDavid.

Figure13:Linearblendingofoctahedronandcube.

7Discussion

Thispaperdescribesanewimplicitsurfacerepresentationbasedonlocalshapefunctions,partitionsofunity,andanoctreehierarchy.StrengthsoftheMulti-LevelPartitionofUnityformulationinclude:

Fastsurfacereconstructionandrendering.Representationofsharpfeatures.Reconstructionfromincompletedata.

Choiceofeitherapproximationorinterpolationofthedataandtheabilitytoadaptivelyvarytheapproximationaccuracy.

GivenapointsetmodelprocessedbytheMPUmethodwithaspec-i edaccuracy,thecomputationaltimeandmemoryusagedependsonthegeometriccomplexityofthemodel:thehigherthegeomet-riccomplexity,thedeepertheoctreeissubdivided.ThisisclearlydemonstratedbyFigure1wherethereconstructionof nefeaturesrequiresadeepersubdivision.

Table1presentsRAMmemoryusageandcomputationaltimemeasurementsforsimultaneousgeneratingandpolygonizingvari-ouspointsetmodels.Notethatourmethodisquitefast.Ourexper-imentswithstate-of-the-artRBF-based3DsurfacereconstructiontechniquessuchasFastRBF[Carretal.2001]andotherssuggestthattheMPUmethodisconsiderablyfasterthantheseothertech-niques.2

Model

NumberofpointsRelativeerrorPeakRAMNumberoftrianglesComp.time

BunnyBunnyscansDragonBuddhaDavid(2mm)34,611362,272433,375543,6254,124,4542.5×101.0×10 38.0×10 4

0.01.0×10 434MB110MB195MB442

Multi-level Partition of Unity Implicits图形学

MB

Multi-level Partition of Unity Implicits图形学

810MB91,104219,676819,704648,332

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

Multi-level Partition of Unity Implicits图形学

1,296,5220:071:461:396:53

Multi-level Partition of Unity Implicits图形学

10:33

Table1:Memoryandcomputationaltimemeasurementsforgenera-tion+polygonizationofMPUimplicitsforvariouspointsetmodels.Computationswereperformedona1.6GHzMobilePentium4with1GBRAM,andtimingsarelistedasminutes:seconds.

2ComparingtheresultsofTable1withthoseofTable2in[Carretal.2001]one

can ndthattheMPUmethodis20-30timesfasterthantheFastRBFtechnique[Carretal.2001].

NoticethatprocessingtimefortheBuddhamodelismorethanthreetimeslongerthenthatforthedragonmodelwhichhasasim-ilarsize.ThisisbecausewereconstructtheBuddhabytheMPUinterpolationwhichrequiresadeepersubdivisionandwidersupportforthecorrespondingpartitionofunityfunctions.

Becauseofitslocalnature,theMPUmethodismoresensi-tivetothequalityofinputdata,especiallythe eldofnormals,tocomparewiththeapproximationandinterpolationtechniquesbasedonglobally-supportedRBFs[Carretal.2001;TurkandO’Brien2002].Nevertheless,accordingtoourexperiments,theMPUmethodissuf cientforaccurateshapereconstructionfromawidevarietyofdatasets.Theparametersinourcurrentimple-mentationoftheMPUapproachareadjustedforprocessingtypicaloutputsoflaserscannerdevices.Webelievethattheparametermodi cationsneededfordifferentclassesofinputarefairlyintu-itiveinordertohandlescattereddataoflower(higher)qualityattheexpenseoflower(higher)computationalspeed.

Unlikethecrustapproach[Amentaetal.1998],theMPUmethodisnotsupportedbyrigorousresultsguaranteeingcorrectrecon-structionofinputdatasatisfyingcertainpropertiesdescribedmath-ematically.Itisapricewepayforahighspeedofourmethod.WewouldlikealsotostressherethatourmethodisnotanRBFtechnique.RBFisaglobalapproximation/interpolationmethodbe-causeofitsglobalvariationalnature.Ourmethodisalocalone.Wemakeuseoftwofunctions,thepartition-of-unityweightsandthelocalpiecewisequadraticapproximationfunctions,whichisdiffer-entthanthesinglefunctionusedbyanRBFapproach.Thistwo-functionapproachgivesbene tssuchassharpfeaturereconstruc-tionthat,todate,havenotbeenpossibleusingRBFs.

SmoothnesspropertiesoftheMPUimplicitsaredeterminedbythoseofweightfunctions(3).ChoosingsmootherweightfunctionswillproducesmootherMPUimplicits.

Similartootherimplicitfunctionshaperepresentationschemes,theMPUimplicitsarenotcapableofrepresentingcorrectlysurfaceswithboundaries.

Weseeanumberofopportunitiestoimproveourapproach.Otherestimationofthedistancefunctionmightbebene cial.Thedistancefunctionisaruledsurfacewithsingularities,thereforeus-ingquadraticfunctionstoapproximatethedistancefunctionisnotoptimal.Aricherlibraryoflocalshapeapproximationscouldbegeneratedinordertoreconstructaccuratelymorecomplexsharpfeatures.Finally,theMPUapproachshouldbewellsuitedtoanout-of-coreimplementationduetothelocalnatureoftheweightingfunctions.

上一页第7页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email