ebook img

Computer vision using local binary patterns PDF

224 Pages·2011·5.005 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 Computer vision using local binary patterns

Computer Vision Using Local Binary Patterns Computational Imaging and Vision ManagingEditor MAXVIERGEVER UtrechtUniversity,Utrecht,TheNetherlands SeriesEditors GUNILLABORGEFORS,CentreforImageAnalysis,SLU,Uppsala,Sweden RACHIDDERICHE,INRIA,SophiaAntipolis,France THOMASS.HUANG,UniversityofIllinois,Urbana,USA KATSUSHIIKEUCHI,TokyoUniversity,Tokyo,Japan TIANZIJIANG,InstituteofAutomation,CAS,Beijing,China REINHARDKLETTE,UniversityofAuckland,Auckland,NewZealand ALESLEONARDIS,ViCoS,UniversityofLjubljana,Ljubljana,Slovenia HEINZ-OTTOPEITGEN,CeVis,Bremen,Germany JOHNK.TSOTSOS,YorkUniversity,Toronto,Canada This comprehensive book series embraces state-of-the-art expository works and advanced researchmonographsonanyaspectofthisinterdisciplinaryfield. Topicscoveredbytheseriesfallinthefollowingfourmaincategories: • ImagingSystemsandImageProcessing • ComputerVisionandImageUnderstanding • Visualization • ApplicationsofImagingTechnologies Onlymonographsormulti-authoredbooksthathaveadistinctsubjectarea,thatiswhereeach chapterhasbeeninvitedinordertofulfillthispurpose,willbeconsideredfortheseries. Volume40 Forfurthervolumes: www.springer.com/series/5754 Matti Pietikäinen (cid:2) Abdenour Hadid (cid:2) Guoying Zhao (cid:2) Timo Ahonen Computer Vision Using Local Binary Patterns MattiPietikäinen GuoyingZhao MachineVisionGroup MachineVisionGroup DepartmentofComputerScienceand DepartmentofComputerScienceand Engineering Engineering UniversityofOulu UniversityofOulu POBox4500 POBox4500 90014Oulu 90014Oulu Finland Finland [email protected].fi [email protected].fi AbdenourHadid TimoAhonen MachineVisionGroup NokiaResearchCenter DepartmentofComputerScienceand PaloAlto,CA Engineering USA UniversityofOulu [email protected] POBox4500 90014Oulu Finland [email protected].fi ISSN1381-6446 ISBN978-0-85729-747-1 e-ISBN978-0-85729-748-8 DOI10.1007/978-0-85729-748-8 SpringerLondonDordrechtHeidelbergNewYork BritishLibraryCataloguinginPublicationData AcataloguerecordforthisbookisavailablefromtheBritishLibrary LibraryofCongressControlNumber:2011932161 MathematicsSubjectClassification: 68T45,68H35,68U10,68T10,97R40 ©Springer-VerlagLondonLimited2011 Apartfromanyfairdealingforthepurposesofresearchorprivatestudy,orcriticismorreview,asper- mittedundertheCopyright,DesignsandPatentsAct1988,thispublicationmayonlybereproduced, storedortransmitted,inanyformorbyanymeans,withthepriorpermissioninwritingofthepublish- ers,orinthecaseofreprographicreproductioninaccordancewiththetermsoflicensesissuedbythe CopyrightLicensingAgency.Enquiriesconcerningreproductionoutsidethosetermsshouldbesentto thepublishers. Theuseofregisterednames,trademarks,etc.,inthispublicationdoesnotimply,evenintheabsenceofa specificstatement,thatsuchnamesareexemptfromtherelevantlawsandregulationsandthereforefree forgeneraluse. Thepublishermakesnorepresentation,expressorimplied,withregardtotheaccuracyoftheinformation containedinthisbookandcannotacceptanylegalresponsibilityorliabilityforanyerrorsoromissions thatmaybemade. Coverdesign:deblik Printedonacid-freepaper SpringerispartofSpringerScience+BusinessMedia(www.springer.com) Preface Humansreceivethegreatmajorityofinformationabouttheirenvironmentthrough sight,andatleast50%ofthehumanbrainisdedicatedtovision.Visionisalsoakey componentforbuildingartificialsystemsthatcanperceiveandunderstandtheiren- vironment.Computervisionislikelytochangesocietyinmanyways;forexample, itwillimprovethesafetyandsecurityofpeople,itwillhelpblindpeoplesee,andit willmakehuman-computerinteractionmorenatural.Withcomputervisionitispos- sibletoprovidemachineswithanabilitytounderstandtheirsurroundings,control thequalityofproductsinindustrialprocesses,helpdiagnosediseasesinmedicine, recognizehumansandtheiractions,andsearchforinformationfromdatabasesus- ingimageorvideocontent. Texture is an important characteristic of many types of images. It can be seen inimagesrangingfrommultispectralremotelysenseddatatomicroscopicimages. Atexturedareainanimagecanbecharacterizedbyanonuniformorvaryingspatial distribution of intensity or color. The variation reflects some changes in the scene being imaged. For example, an image of mountainous terrain appears textured. In outdoorimages,trees,bushes,grass,sky,lakes,roads,buildingsetc.appearasdif- ferenttypesoftexture.Thespecificstructureofthetexturedependsonthesurface topography and albedo, the illumination of the surface, and the position and fre- quency response of the viewer. An X-ray of diseased tissue may appear textured duetothedifferentabsorptioncoefficientsofhealthyanddiseasedcellswithinthe tissue. Texturecanplayakeyroleinawidevarietyofapplicationsofcomputervision. Thetraditionalareasofapplicationconsideredfortextureanalysisincludebiomedi- calimageanalysis,industrialinspection,analysisofsatelliteoraerialimagery,doc- umentimageanalysis,andtexturesynthesisforcomputergraphicsoranimation. Texture analysis has been a topic of intensive research since the 1960s, and a widevarietyoftechniquesfordiscriminatingtextureshavebeenproposed.Mostof theproposedmethodshavenotbeen,however,capabletoperformwellenoughfor real-world textures and are computationally too complex to meet the real-time re- quirementsofmanyapplications.Inrecentyears,verydiscriminativeandcomputa- tionallyefficientlocaltexturedescriptorshavebeendeveloped,suchaslocalbinary v vi Preface patterns(LBP),whichhasledtoasignificantprogressinapplyingtexturemethods tovariouscomputervisionproblems.Thefocusoftheresearchhasbroadenedfrom 2Dtexturesto3Dtexturesandspatiotemporal(dynamic)textures. With this progress the emerging application areas of texture analysis will also cover such modern fields as face analysis and biometrics, object recognition, mo- tion analysis, recognition of actions, content-based retrieval from image or video databases,andvisualspeechrecognition.Thisbookprovidesanexcellentoverview how texture methods can be used for solving these kinds of problems, as well as moretraditionalapplications.EspeciallytheuseofLBPinbiomedicalapplications andbiometricrecognitionsystemshasgrownrapidlyinrecentyears. Thelocalbinarypattern(LBP)isasimpleyetveryefficientoperatorwhichlabels thepixelsofanimagebythresholdingtheneighborhoodofeachpixelandconsiders theresultasabinarynumber.TheLBPmethodcanbeseenasaunifyingapproach to the traditionally divergent statistical and structural models of texture analysis. Perhaps the most important property of the LBP operator in real-world applica- tionsisitsinvarianceagainstmonotonicgraylevelchangescaused,forexample,by illumination variations. Another equally important is its computational simplicity, whichmakesitpossibletoanalyzeimagesinchallengingreal-timesettings.LBPis alsoveryflexible:itcanbeeasilyadaptedtodifferenttypesofproblemsandused togetherwithotherimagedescriptors. The book is divided into five parts. Part I provides an introduction to the book contentsandanin-depthdescriptionofthelocalbinarypatternoperator.Acompre- hensivesurveyofdifferentvariantsofLBPisalsopresented.PartIIdealswiththe analysisofstillimagesusingLBPoperators.Applicationsintextureclassification, segmentation,descriptionofinterestregions,content-basedimageretrievaland3D recognitionoftexturedsurfacesareconsidered.ThetopicofPartIIIismotionanaly- sis,withapplicationsindynamictexturerecognitionandsegmentation,background modelinganddetectionofmovingobjects,andrecognitionofactions.PartIVdeals withfaceanalysis.TheLBPoperatorsareusedforanalyzingstillimagesandimage sequences. The specific application problem of visual speech recognition is pre- sentedinmoredetail.Finally,PartVprovidesanintroductiontosomerelatedwork bydescribingrepresentativeexamplesofusingLBPindifferentapplications,such asbiometrics,visualinspectionandbiomedicalapplications,forexample. Wewouldliketothankallco-authorsofourLBPpapersfortheirinvaluablecon- tributionstothecontentsofthisbook.Firstofall,specialthankstoTimoOjalaand David Harwood who started LBP investigations in our group in fall 1992 during DavidHarwood’svisitfromtheUniversityofMarylandtoOulu.SincethenTimo OjalamademanycentralcontributionstoLBPuntil2002whenourveryfrequently cited paper was published in IEEE Transactions on Pattern Analysis and Machine Intelligence. Topi Mäenpää played also a very significant role in many develop- mentsofLBP.Otherkeycontributors,inalphabeticorder,includeJieChen,Xiaoyi Feng,YimoGuo,ChuHe,MarkoHeikkilä,ViliKellokumpu,StanZ.Li,JiriMatas, TomiNurmela,CordeliaSchmid,MattiTaini,ValtteriTakala,andMarkusTurtinen. We also thank the anonymous reviewers, whose constructive comments helped us improvethebook. Preface vii MatlabandCcodesofthebasicLBPoperatorsandsomevideodemonstrations canbefoundfromanaccompanyingwebsiteatwww.cse.oulu.fi/MVG/LBP_Book. ForabibliographyofLBP-relatedresearchandlinkstomanypapers,seewww.cse. oulu.fi/MVG/LBP_Bibliography. Oulu,Finland MattiPietikäinen Oulu,Finland AbdenourHadid Oulu,Finland GuoyingZhao PaloAlto,CA TimoAhonen Contents PartI LocalBinaryPatternOperators 1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.1 TheRoleofTextureinComputerVision . . . . . . . . . . . . . . 3 1.2 MotivationandBackgroundforLBP . . . . . . . . . . . . . . . . 4 1.3 ABriefHistoryofLBP . . . . . . . . . . . . . . . . . . . . . . . 6 1.4 OverviewoftheBook . . . . . . . . . . . . . . . . . . . . . . . . 7 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 2 LocalBinaryPatternsforStillImages . . . . . . . . . . . . . . . . . 13 2.1 BasicLBP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 2.2 DerivationoftheGenericLBPOperator . . . . . . . . . . . . . . 13 2.3 MappingsoftheLBPLabels:UniformPatterns. . . . . . . . . . . 16 2.4 RotationalInvariance . . . . . . . . . . . . . . . . . . . . . . . . 18 2.4.1 RotationInvariantLBP . . . . . . . . . . . . . . . . . . . 19 2.4.2 RotationInvarianceUsingHistogramTransformations . . . 20 2.5 ComplementaryContrastMeasure . . . . . . . . . . . . . . . . . 21 2.6 Non-parametricClassificationPrinciple . . . . . . . . . . . . . . . 23 2.7 MultiscaleLBP . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 2.8 Center-SymmetricLBP . . . . . . . . . . . . . . . . . . . . . . . 25 2.9 OtherLBPVariants . . . . . . . . . . . . . . . . . . . . . . . . . 26 2.9.1 Preprocessing . . . . . . . . . . . . . . . . . . . . . . . . 26 2.9.2 NeighborhoodTopology . . . . . . . . . . . . . . . . . . . 31 2.9.3 ThresholdingandEncoding . . . . . . . . . . . . . . . . . 32 2.9.4 MultiscaleAnalysis . . . . . . . . . . . . . . . . . . . . . 35 2.9.5 HandlingRotation . . . . . . . . . . . . . . . . . . . . . . 37 2.9.6 HandlingColor . . . . . . . . . . . . . . . . . . . . . . . 38 2.9.7 FeatureSelectionandLearning . . . . . . . . . . . . . . . 39 2.9.8 ComplementaryDescriptors . . . . . . . . . . . . . . . . . 42 2.9.9 OtherMethodsInspiredbyLBP . . . . . . . . . . . . . . . 42 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 ix

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.