ebook img

Geometric Curve Evolution and Image Processing PDF

187 Pages·2003·2.219 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Geometric Curve Evolution and Image Processing

1805 Lecture Notes in Mathematics Editors: J.–M.Morel,Cachan F.Takens,Groningen B.Teissier,Paris 3 Berlin Heidelberg NewYork HongKong London Milan Paris Tokyo Fre´de´ric Cao Geometric Curve Evolution and Image Processing 1 3 Author Fre´de´ricCao IRISA/INRIA CampusUniversitairedeBeaulieu 35042RennesCedex France e-mail:[email protected] http://www.irisa.fr/prive/fcao Cataloging-in-PublicationDataappliedfor BibliographicinformationpublishedbyDieDeutscheBibliothek DieDeutscheBibliothekliststhispublicationintheDeutscheNationalbibliografie; detailedbibliographicdataisavailableintheInternetathttp://dnb.ddb.de MathematicsSubjectClassification(2000):35K65,35D05,53A04,68U10 ISSN0075-8434 ISBN3-540-00402-5Springer-VerlagBerlinHeidelbergNewYork Thisworkissubjecttocopyright.Allrightsarereserved,whetherthewholeorpartofthematerialis concerned,specificallytherightsoftranslation,reprinting,reuseofillustrations,recitation,broadcasting, reproductiononmicrofilmorinanyotherway,andstorageindatabanks.Duplicationofthispublication orpartsthereofispermittedonlyundertheprovisionsoftheGermanCopyrightLawofSeptember9,1965, initscurrentversion,andpermissionforusemustalwaysbeobtainedfromSpringer-Verlag.Violationsare liableforprosecutionundertheGermanCopyrightLaw. Springer-VerlagBerlinHeidelbergNewYorkamemberofBertelsmannSpringer Science+BusinessMediaGmbH http://www.springer.de ©Springer-VerlagBerlinHeidelberg2003 PrintedinGermany Theuseofgeneraldescriptivenames,registerednames,trademarks,etc.inthispublicationdoesnotimply, evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevantprotectivelaws andregulationsandthereforefreeforgeneraluse. Typesetting:Camera-readyTEXoutputbytheauthor SPIN:10904251 41/3142/du-543210-Printedonacid-freepaper Preface These lectures intend to give a self-contained exposure of some techniques for computingtheevolutionofplanecurves.Themotionsofinterestaretheso-called motionsbycurvature.Theymeanthat,atanyinstant,eachpointofthecurvemoves withanormalvelocityequaltoafunctionofthecurvatureatthispoint.Thiskindof evolutionisofsomeinterestindifferentialgeometry,forinstanceintheproblemof minimalsurfaces.Theinterestisnotonlytheoreticalsincethemotionsbycurvature appearinthemodelingofvariousphenomenaascrystalgrowth,flamepropagation andinterfacesbetweenphases.Morerecently,theseequationshavealsoappearedin theyoungfieldofimageprocessingwheretheyprovideanefficientwaytosmooth curvesrepresentingthecontoursoftheobjects.Thissmoothingisanecessarystepfor imageanalysisassoonastheanalysisusessomelocalcharacteristicsofthecontours. Indeed, natural images are very noisy and differential features are unreliable if one is not careful before computing them. A solution consists in smoothing the curvestoeliminatethesmalloscillationswithoutchangingtheglobalshapeofthe contours.Whatkindofsmoothingissuitableforsuchatask?Theanswershallbe givenbyanaxiomaticapproachwhoseconclusionsarethattheclassofadmissible motionsisreducedtothemotionsbycurvature.Oncethisisestablished,thewell- posedness of these equations has to be examined. For certain particular motions, thisturnstobetruebutnocompleteresultsareavailableforthegeneralexistenceof thesemotions.Thisproblemshallbeturnedaroundbyintroducingaweaknotion of solution using the theory of viscosity solutions of partial differential equations (PDE).Acompletetheoryofexistenceanduniquenessofthoseequationswillbe presented,asself-containedaspossible.(Onlyatechnical,thoughimportant,lemma willbeskipped.)Thenumericalresolutionofthemotionsbycurvatureisthenext topic of interest. After a rapid review of the most commonly used algorithms, a completelydifferentnumericalschemeispresented.Itsoriginalityisthatitsatisfies exactlythesameinvariancepropertiesastheequationsofmotionbycurvature.It is also inconditionally stable and its convergence can be proved in the sense of viscositysolutions.Moreover,itallowstopreciselycomputemotionsbycurvature, when the normal velocity is a power of the curvature more than 3, or even 10 in VI Preface some cases, which seems a priori nearly impossible in a numerical point of view. Manynumericalexperimentsarepresented. Whothisvolumeisaddressedto? We hope that these notes shall interest people from both communities of applied mathematics and image processing. We tried to make them as self-contained as possible.Nevertheless,weskippedthemostdifficultresultssincetheirproofuses techniquesthatwouldhaveledustoofarfromourmainway.Indeed,theselectures areaddressedtoresearchersdiscoveringthecommonfieldofmathematicsandim- ageprocessingbutalsotograduateandPhDstudentswantingtospanatheoryfrom AtoZ:fromthebasicaxioms,tomathematicalresultsandnumericalapplications. ThechaptersaremostlyindependentexceptChap.6thatusesresultsfromChap.4. Thebibliographyoneverysubjectwetackleishuge,andwecannotpretendtogive exhaustive references on differential geometry, viscosity solutions, mathematical morphologyorscalespacetheory.Attheendofmostchapters,wegiveshortbibli- ographicalnotesdetailinginafewwordsthemainstepsthatproducedsignificant advancesinthetheory. Acknowledgements MyfirstthanksgotoJean-MichelMorelwhowasmyadvisorduringmythesisand also proposed me to write these notes. It has been a constant pleasure to work in histeaminwhichcollaborationsareasnumerousasfruitful.Iwouldliketothank himforhisscientificsupportbutalsoforalltheotherqualitieshedevelopsinhis relations. IwouldalsoliketoexpressmygratitudetoYvesMeyer,withwhomanydiscussion isalwaysanopportunitytolearnsomethingandtorealizewehavestillsomuchto learn.Ihavealsobeenalwaysimpressedbyhisenthusiasm,andthankhimsincerely forhisencouragements. IowealsomuchtoLionelMoisan,whocontributedtotheredactionofChap.6and whosehelp,speciallyforthenumericalpart,wasveryprecious. ManythanksalsotoJean-PierreLeCadreforhiscarefulreadingofthistext. IwouldalsoliketothankAndre´sAlmansa,VicentCaselles,FrancineCatte´,Agne`s Desolneux,Franc¸oiseDibos,JacquesFroment,YannGousseau,Fre´de´ricGuichard, Georges Koepfler, Franc¸ois Malgouyres, Simon Masnou, Pascal Monasse, Pablo Muse´,DenisPasquignon,Fre´de´ricSurforallthediscussionswehadforthelastfew years. Rennes,October2002 Fre´de´ricCao Contents Preface ....................................................... V PartI Thecurvesmoothingproblem 1 Curveevolutionandimageprocessing ......................... 3 1.1 Shaperecognition ......................................... 5 1.1.1 Axiomsforshaperecognition... ....................... 7 1.1.2 ...andtheirconsequences ............................ 8 1.2 Curvesmoothing.......................................... 8 1.2.1 Thelinearcurvescalespace .......................... 9 1.2.2 Towardsanintrinsicheatequation ..................... 10 1.3 Anaxiomaticapproachofcurveevolution ..................... 12 1.3.1 Basicrequirements.................................. 12 1.3.2 Firstconclusionsandfirstmodels...................... 13 1.4 Imageandcontoursmoothing ............................... 15 1.5 Applications.............................................. 15 1.5.1 Activecontours..................................... 15 1.5.2 Principlesofashaperecognitionalgorithm.............. 17 1.5.3 Opticalcharacterrecognition ......................... 18 1.6 Organizationofthevolume ................................. 19 1.7 Bibliographicalnotes ...................................... 20 2 Rudimentarybasesofcurvegeometry ......................... 23 2.1 Jordancurves............................................. 23 2.2 Lengthofacurve.......................................... 24 2.3 Euclideanparameterization ................................. 25 2.4 Motionofgraphs.......................................... 27 VIII Contents PartII Theoreticalcurveevolution 3 Geometriccurveshorteningflow.............................. 31 3.1 Whatkindofequationsforcurvesmoothing? .................. 32 3.1.1 Invariantflows ..................................... 32 3.1.2 Symmetrygroupofflow ............................. 32 3.2 Differentialinvariants ...................................... 40 3.2.1 Generalformofinvariantflows ....................... 40 3.2.2 ThemeancurvatureflowistheEuclideanintrinsicheatflow 42 3.2.3 Theaffineinvariantflow:thesimplestaffineinvariant curveflow ......................................... 42 3.3 Generalpropertiesofsecondorderparabolicflows:adigest ...... 46 3.3.1 Existenceanduniquenessformeancurvatureandaffine flows.............................................. 46 3.3.2 Short-timeexistenceinthegeneralcase ................ 48 3.3.3 Evolutionofconvexcurves ........................... 48 3.3.4 Evolutionofthelength .............................. 49 3.3.5 Evolutionofthearea ................................ 49 3.4 Smoothingstaircases....................................... 50 3.5 Bibliographicalnotes ...................................... 53 4 Curveevolutionandlevelsets ................................ 55 4.1 Fromcurveoperatorstofunctionoperatorsandviceversa........ 57 4.1.1 Signeddistancefunctionandsupportingfunction ........ 57 4.1.2 Monotoneandtranslationinvariantoperators ............ 57 4.1.3 Levelsetsandtheirproperties......................... 58 4.1.4 Extensionofsetsoperatorstofunctionsoperators ........ 60 4.1.5 Characterizationofmonotone,contrastinvariantoperator . 62 4.1.6 Asymptoticbehaviorofmorphologicaloperators......... 64 4.1.7 MorphologicaloperatorsyieldPDEs ................... 69 4.2 CurveevolutionandScaleSpacetheory....................... 70 4.2.1 MultiscaleanalysisaregivenbyPDEs ................. 70 4.2.2 Morphologicalscalespace ........................... 73 4.3 Viscositysolutions......................................... 74 4.3.1 Definitionofviscositysolution........................ 76 4.3.2 Proofofuniqueness:themaximumprinciple ............ 80 4.3.3 ExistenceofsolutionbyPerron’sMethod............... 85 4.3.4 Contrastinvarianceoflevelsetsflow................... 88 4.3.5 Viscositysolutionsshortenlevellines .................. 89 4.4 Morphologicaloperatorsandviscositysolution................. 90 4.4.1 Medianfilterandmeancurvaturemotion ............... 90 4.4.2 Affineinvariantschemes ............................. 92 4.5 Conclusions .............................................. 93 4.6 Curvaturethresholding ..................................... 93 Contents IX 4.6.1 Viscosityapproach .................................. 93 4.6.2 Openingandclosing................................. 95 4.7 Alternativeweaksolutionsofcurveevolution .................. 100 4.7.1 Brakke’svarifoldsolution ............................ 100 4.7.2 Reactiondiffusionapproximation...................... 100 4.7.3 Minimalbarriers.................................... 101 4.8 Bibliographicalnotes ...................................... 101 PartIII Numericalcurveevolution 5 Classicalnumericalmethodsforcurveevolution ................ 107 5.1 Parametricmethods........................................ 107 5.1.1 Finitedifferencemethods ............................ 107 5.1.2 Finiteelementschemes .............................. 108 5.2 Nonparametricmethods.................................... 108 5.2.1 Sethian’slevelsetsmethods .......................... 108 5.2.2 AlvarezandGuichard’sfinitedifferencesscheme ........ 109 5.2.3 Amonotoneandconvergentfinitedifferenceschemes..... 109 5.2.4 Bence,MerrimanandOsherschemeformeancurvature motion ............................................ 109 5.2.5 Ellipticregularization................................ 110 6 Ageometricalschemeforcurveevolution ...................... 111 6.1 Preliminarydefinitions ..................................... 112 6.2 Erosion .................................................. 114 6.3 Propertiesoftheerosion.................................... 115 6.4 Erosionandlevelsets ...................................... 119 6.4.1 Consistency........................................ 119 6.4.2 Convergence ....................................... 124 6.5 Evolutionbyapowerofthecurvature ........................ 124 6.5.1 Consistency........................................ 125 6.5.2 Convergence ....................................... 129 6.6 Anumericalimplementationoftheerosion .................... 133 6.6.1 Scalecovariance.................................... 133 6.6.2 Generalalgorithm................................... 135 6.6.3 Erodedsetandenvelope ............................. 136 6.6.4 Swallowtails....................................... 138 6.7 Numericalexperiments..................................... 139 6.7.1 Evolvingcircles .................................... 140 6.7.2 Convexpolygons ................................... 141 6.7.3 Unclosedcurve..................................... 142 6.7.4 Evolvingnonconvexcurves .......................... 142 6.7.5 Invariance ......................................... 150 6.7.6 Numericalmaximumprinciple ........................ 152 X Contents 6.7.7 Imagefiltering...................................... 156 6.8 Bibliographicalnotes ...................................... 166 Conclusionandperspectives ..................................... 167 Discussion .................................................... 167 Openproblems ................................................ 169 A ProofofThm.4.34.......................................... 171 References .................................................... 177 Index......................................................... 185

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.