ImageProcessing1-Introduction 数字图像处理 英文版

CourseWebsite::[e-mail]2of36Introduction“Onepictureisworthmorethantenthousandwords”Anonymous3of36MiscellaneaLectures:–Thursdays12:00–13:00–Fridays15:00–16:00Labs:–Wednesdays09:00–11:00WebSite:–Previousyear’sslidesareavailablehere–SlidesetcwillalsobeavailableonWebCTE-mail:[e-mail]4of36References“DigitalImageProcessing”,RafaelC.Gonzalez&RichardE.Woods,Addison-Wesley,2002–Muchofthematerialthatfollowsistakenfromthisbook“MachineVision:AutomatedVisualInspectionandRobotVision”,DavidVernon,PrenticeHall,1991–Availableonlineat:homepages.inf.ed.ac.uk/rbf/BOOKS/VERNON/5of36ContentsThislecturewillcover:–Whatisadigitalimage?–Whatisdigitalimageprocessing?–Historyofdigitalimageprocessing–Stateoftheartexamplesofdigitalimageprocessing–Keystagesindigitalimageprocessing6of36WhatisaDigitalImage?Adigitalimageisarepresentationofatwo-dimensionalimageasafinitesetofdigitalvalues,calledpictureelementsorpixelsImagestakenfromGonzalez&Woods,DigitalImageProcessing(2002)7of36WhatisaDigitalImage?(cont…)Pixelvaluestypicallyrepresentgraylevels,colours,heights,opacitiesetcRememberdigitizationimpliesthatadigitalimageisanapproximationofarealscene1pixelImagestakenfromGonzalez&Woods,DigitalImageProcessing(2002)8of36WhatisaDigitalImage?(cont…)Commonimageformatsinclude:–1sampleperpoint(B&WorGrayscale)–3samplesperpoint(Red,Green,andBlue)–4samplesperpoint(Red,Green,Blue,and“Alpha”,a.k.a.Opacity)Formostofthiscoursewewillfocusongrey-scaleimages9of36WhatisDigitalImageProcessing?Digitalimageprocessingfocusesontwomajortasks–Improvementofpictorialinformationforhumaninterpretation–Processingofimagedataforstorage,transmissionandrepresentationforautonomousmachineperceptionSomeargumentaboutwhereimageprocessingendsandfieldssuchasimageanalysisandcomputervisionstart10of36WhatisDIP?(cont…)Thecontinuumfromimageprocessingtocomputervisioncanbebrokenupintolow-,mid-andhigh-levelprocessesLowLevelProcessInput:ImageOutput:ImageExamples:Noiseremoval,imagesharpeningMidLevelProcessInput:ImageOutput:AttributesExamples:Objectrecognition,segmentationHighLevelProcessInput:AttributesOutput:UnderstandingExamples:Sceneunderstanding,autonomousnavigationInthiscoursewewillstophere11of36HistoryofDigitalImageProcessingEarly1920s:Oneofthefirstapplicationsofdigitalimagingwasinthenews-paperindustry–TheBartlanecablepicturetransmissionservice–ImagesweretransferredbysubmarinecablebetweenLondonandNewYork–PictureswerecodedforcabletransferandreconstructedatthereceivingendonatelegraphprinterEarlydigitalimageImagestakenfromGonzalez&Woods,DigitalImageProcessing(2002)12of36HistoryofDIP(cont…)Midtolate1920s:ImprovementstotheBartlanesystemresultedinhigherqualityimages–Newreproductionprocessesbasedonphotographictechniques–IncreasednumberoftonesinreproducedimagesImproveddigitalimageEarly15tonedigitalimageImagestakenfromGonzalez&Woods,DigitalImageProcessing(2002)13of36HistoryofDIP(cont…)1960s:Improvementsincomputingtechnologyandtheonsetofthespaceraceledtoasurgeofworkindigitalimageprocessing–1964:ComputersusedtoimprovethequalityofimagesofthemoontakenbytheRanger7probe–SuchtechniqueswereusedinotherspacemissionsincludingtheApollolandingsApictureofthemoontakenbytheRanger7probeminutesbeforelandingImagestakenfromGonzalez&Woods,DigitalImageProcessing(2002)14of36HistoryofDIP(cont…)1970s:Digitalimageprocessingbeginstobeusedinmedicalapplications–1979:SirGodfreyN.Hounsfield&Prof.AllanM.CormacksharetheNobelPrizeinmedicinefortheinventionoftomography,thetechnologybehindComputerisedAxialTomography(CAT)scansTypicalheadsliceCATimageImagestakenfromGonzalez&Woods,DigitalImageProcessing(2002)15of36HistoryofDIP(cont…)1980s-Today:Theuseofdigitalimageprocessingtechniqueshasexplodedandtheyarenowusedforallkindsoftasksinallkindsofareas–Imageenhancement/restoration–Artisticeffects–Medicalvisualisation–Industrialinspection–Lawenforcement–Humancomputerinterfaces16of36Examples:ImageEnhancementOneofthemostcommonusesofDIPtechniques:improvequality,removenoiseetcImagestakenfromGonzalez&Woods,DigitalImageProcessing(2002)17of36Examples:TheHubbleTelescopeLaunchedin1990theHubbletelescopecantakeimagesofverydistantobjectsHowever,anincorrectmirrormademanyofHubble’simagesuselessImageprocessingtechniqueswereusedtofixthis18of36Examples:ArtisticEffectsArtisticeffectsareusedtomakeimagesmorevisuallyappealing,toaddspecialeffectsandtomakecompositeimages19of36Examples:MedicineTakeslicefromMRIscanofcanineheart,andfindboundariesbetweentypesoftissue–Imagewithgraylevelsrepresentingtissuedensity–UseasuitablefiltertohighlightedgesOriginalMRIImageofaDogHeartEdgeDetectionImageImagestakenfromGonzalez&Woods,DigitalImageProcessing(2002)20of36Examples:GISGeographicInformationSystems–Digitalimageprocessingtechniquesareusedextensivelytomanipulatesatelliteimagery–Terrainclassification–MeteorologyImagestakenfromGonzalez&Woods,DigitalImageProcessing(2002)21of36Examples:GIS(cont…)Night-TimeLightsoftheWorlddataset–Globalinventoryofhumanset