.searchofdigitalimageprocessingtechniqueInterestindigitalimageprocessingmethodsstemsfromtwoprincipalapplicationareas:improvementofpictorialinformationforhumaninterpretation;andprocessingofimagedataforstorage,transmission,andrepresentationforautonomousmachineperception.Thischapterhasseveralobjectives:(1)todefinethescopeofthefieldthatwecallimageprocessing;(2)togiveahistoricalperspectiveoftheoriginsofthisfield;(3)togiveanideaofthestateoftheartinimageprocessingbyexaminingsomeoftheprincipalareainwhichitisapplied;(4)todiscussbrieflytheprincipalapproachesusedindigitalimageprocessing;(5)togiveanoverviewofthecomponentscontainedinatypical,general-purposeimageprocessingsystem;and(6)toprovidedirectiontothebooksandotherliteraturewhereimageprocessingworknormallyisreporter.1.1WhatIsDigitalImageProcessing?Animagemaybedefinedasatwo-dimensionalfunction,f(x,y),wherexandyarespatial(plane)coordinates,andtheamplitudeoffatanypairofcoordinates(x,y)iscalledtheintensityorgrayleveloftheimageatthatpoint.Whenx,y,anddigitalimage.Thefieldofdigitalimageprocessingreferstoprocessingdigitalimagesbymeansofadigitalcomputer.Notethatadigitalimageiscomposedofafinitenumberofelements,eachofwhichhasaparticularlocationandvalue.Theseelementsarereferredtoaspictureelements,imageelements,pels,andpixels.Pixelisthetermmostwidelyusedtodenotetheelementsofadigitalimage.WeconsiderthesedefinitionsinmoreformaltermsinChapter2.Visionisthemostadvancedofoursenses,soitisnotsurprisingthatimagesplaythesinglemostimportantroleinhumanperception.However,unlikehumanwhoarelimitedtothevisualbandoftheelectromagnetic(EM)spectrum,imagingmachinescoveralmosttheentireEMspectrum,rangingfromgammatoradiowaves.Theycanoperateonimagesgeneratedbysourcesthathumanarenotaccustomedtoassociatingwithimage.Theseincludeultrasound,electronmicroscopy,andcomputer-generatedimages.Thus,digitalimageprocessingencompassesawideand.variedfieldofapplication.Thereisnogeneralagreementamongauthorsregardingwhereimageprocessingstopsandotherrelatedareas,suchasimageanalysisandcomputervision,start.Sometimesadistinctionismadebydefiningimageprocessingasadisciplineinwhichboththeinputandoutputofaprocessareimages.Webelievethistobealimitingandsomewhatartificialboundary.Forexample,underthisdefinition,eventhetrivialtaskofcomputingtheaverageintensityofanimage(whichyieldsasinglenumber)wouldnotbeconsideredanimageprocessingoperation.Ontheotherhand,therearefieldssuchascomputervisionwhoseultimategoalistousecomputertoemulatehumanvision,includinglearningandbeingabletomakeinferencesandtakeactionsbasedonvisualinputs.Thisareaitselfisabranchofartificialintelligence(AI)whoseobjectiveistoemulatehumanintelligence.ThisfieldofAIisinitsearlieststagesofinfancyintermsofdevelopment,withprogresshavingbeenmuchslowerthanoriginallyanticipated.Theareaofimageanalysis(alsocalledimageunderstanding)isinbetweenimageprocessingandcomputervision.Therearenoclear-cutboundariesinthecontinuumfromimageprocessingatoneendtocomputervisionattheother.However,oneusefulparadigmistoconsiderthreetypesofcomputerizedprocessesisthiscontinuum:low-,mid-,andhigh-everprocesses.Low-levelprocessesinvolveprimitiveoperationsuchasimagepreprocessingtoreducenoise,contrastenhancement,andimagesharpening.Alow-levelprocessischaracterizedbythefactthatbothitsinputandoutputareimages.Mid-levelprocessingonimagesinvolvestaskssuchassegmentation(partitioninganimageintoregionsorobjects),descriptionofthoseobjectstoreducethemtoaformsuitableforcomputerprocessing,andclassification(recognition)ofindividualobject.Amid-levelprocessischaracterizedbythefactthatitsinputsgenerallyareimages,butitsoutputisattributesextractedfromthoseimages(e.g.,edgescontours,andtheidentityofindividualobject).Finally,higher-levelprocessinginvolves“makingsense”ofanensembleofrecognizedobjects,asinimageanalysis,and,atthefarendofthecontinuum,performingthecognitivefunctionnormallyassociatedwithvision.Basedontheprecedingcomments,weseethatalogicalplaceofoverlapbetweenimageprocessingandimageanalysisistheareaofrecognitionofindividualregionsorobjectsinanimage.Thus,whatwecallinthisbookdigitalimageprocessingencompassesprocesseswhoseinputsandoutputsareimagesand,inaddition,encompassesprocessesthatextractattributesfromimages,uptoand.includingtherecognitionofindividualobjects.Asasimpleillustrationtoclarifytheseconcepts,considertheareaofautomatedanalysisoftext.Theprocessesofacquiringanimageoftheareacontainingthetext.Preprocessingthatimages,extracting(segmenting)theindividualcharacters,describingthecharactersinaformsuitableforcomputerprocessing,andrecognizingthoseindividualcharactersareinthescopeofwhatwecalldigitalimageprocessinginthisbook.Makingsenseofthecontentofthepagemaybeviewedasbeinginthedomainofimageanalysisandevencomputervision,dependingonthelevelofcomplexityimpliedbythestatement“makingcense.”Aswillbecomeevidentshortly,digitalimageprocessing,aswehavedefinedit,isusedsuccessfullyinabroadrangofareasofexceptionalsocialandeconomicvalue.Theconceptsdevelopedinthefollowingchaptersarethefoundationforthemethodsusedinthoseapplicationareas.rocessingOneofthefirstapplicationsofdigitalimageswasinthenewspaperindustry,whenpictureswerefirstsentbysubmarinecablebetweenLondonandNewYork.IntroductionoftheBartlanecablepicturetransmissionsystemintheearly1920sreducedthetimerequiredtotransportapictureacrosstheAtlanticfrommorethanaweektolessthanthreehours.Specializedprintingequipmentcodedpicturesforcabletransmissionandthenreconstructedthematthereceivingend.Figure1.1wastransmittedinthiswayandreproducedonatelegraphprinterfittedwithtypefacessimulatingahalftonepattern.Someoftheinitialproblemsinimprovingthevisualqualityoftheseearlydigitalpictureswererelatedtotheselectionofprintingproceduresandthedistributionofintensitylevels.TheprintingmethodusedtoobtainFig.1.1wasabandonedtowardtheendof1921infavorofatechniquebasedonphotographicreproductionmadefromtapesperforatedatthetelegraphreceivingterminal.Figure1.2showsanimagesobtainedusingthismethod.TheimprovementsoverFig.1.1areevident,bothintonalqualityandinresolution..FIGURE1.1AdigitalpictureproducedinFIGURE1.2Adigitalpicturefromacodedtapebyatelegraphprintermadein1922fromatapepunchedWithspecialtypefaces(McFarlane)afterthesignalshadcrossedtheAtlantictwiceSomeerrorsareVisible(McFarlane)TheearlyBartlanesystemswerecapableofcodingimagesinfivedistinctlevelofgray.Thiscapabilitywasincreasedto15levelsin1929.Figure1.3istypicaloftheimagesthatcouldbeobtainedusingthe15-toneequipment.Duringthisperiod,introductionofasystemfordevelopingafilmplatevialightbeamsthatweremodulatedbythecodedpicturetapeimprovedthereproductionprocessconsiderably.Althoughtheexamplesjustcitedinvolvedigitalimages,theyarenotconsidereddigitalimageprocessingresultsinthecontextofourdefinitionbecausecomputerwerenotinvolvedintheircreation.Thus,thehistoryofdigitalprocessingisintimatelytiedtothedevelopmentofthedigitalcomputer.Infactdigitalimagesrequiresomuchstorageandcomputationalpowerthatprogressinthefieldofdigitalimageprocessinghasbeendependentonthedevelopmentofdigitalcomputersofsupportingtechnologiesthatincludedatastorage,display,andtransmission.TheideaofacomputergoesbacktotheinventionoftheabacusinAsiaMinor,morethan5000yearsago.Morerecently,thereweredevelopmentsinthepasttwocenturiesthatarethefoundationofwhatwecallcomputertoday.However,thebasisforwhatwecallamoderndigitalcomputerdatesbacktoonlythe1940swiththeintroductionbyJohnvonNeumannoftwokeyconcepts:(1)amemorytoholdastoredprogramanddata,and(2)conditionalbranching.Theretwoideasarethefoundationofacentralprocessingunit(CPU),whichisattheheartofcomputertoday.StartingwithvonNeumann,therewereaseriesofadvancesthatledtocomputers.powerfulenoughtobeusedfordigitalimageprocessing.Briefly,theseadvancesmaybesummarizedasfollow:(1)theinventionofthetransistorbyBellLaboratoriesin1948;(2)thedevelopmentinthe1950sand1960softhehigh-levelprogramminglanguagesCOBOL(CommonBusiness-OrientedLanguage)andFORTRANmulaTranslator(3)theinventionoftheintegratedcircuit(IC)atTexasInstrumentsin1958;(4)thedevelopmentofoperatingsystemintheearly1960s;(5)thedevelopmentofthemicroprocessor(asinglechipconsistingofthecentralprocessingunit,memory,andinputandoutputcontrols)byInterintheearly1970s;(6)introductionbyIBMofthepersonalcomputerin1981;(7)progressiveminiaturizationofcomponents,startingwithlargescaleintegration(LI)inthelate1970s,thenverylargescaleintegration(VLSI)inthe1980s,tothepresentuseofultralargescaleintegration(ULSI).Figure3In1929fromLondontoCeneralePershingthatNewYorkdeliverswith15leveltoneequipmentsthroughcablewithFochdonotthephotographbydecorationConcurrentwiththeseadvancesweredevelopmentintheareasofmassstorageanddisplaysystems,bothofwhicharefundamentalrequirementsfordigitalimageprocessing.Thefirstcomputerspowerfulenoughtocarryoutmeaningfulimageprocessingtasksappearedintheearly1960s.Thebirthofwhatwecalldigitalcessingtodaycanbetracedtotheavailabilityofthosemachinesandtheonsetoftheapaceprogramduringthatperiod.Ittookthecombinationofthosetwodevelopmentstobringintofocusthepotentialofdigitalimageprocessingconcepts.WorkonusingcomputertechniquesforimprovingimagesfromaspaceprobebeganattheJetPropulsionLaboratory(Pasadena,California)in1964whenpicturesofthemoontransmittedbyRanger7wereprocessedbyacomputertocorrectvarioustypesofimagedistortioninherentintheon-boardtelevisioncamera.Figure1.4showsthefirstimageofthemoontakenbyRanger7onJuly31,1964at9:09A.M.EasternDaylightTime(EDT),about17minutesbeforeimpactingthelunarsurface(themarkers,calledreseaumark,areusedforgeometriccorrections,asdiscussedinChapter5).ThisalsoisthefirstimageofthemoontakenbyaU.S.spacecraft.Theimaginglessonslearnedwithranger7servedasthebasisforimprovedmethodsusedtoenhanceandrestoreimagesfromtheSurveyormissionstothemoon,theMarinerseriesofflybymissiontoMars,theApollomannedflightstothemoon,andothers.Inparallelwithspaceapplication,digitalimageprocessingtechniquesbeganinthelate1960sandearly1970stobeusedinmedicalimaging,remoteEarthresourcesobservations,andastronomy.Theinventionintheearly1970sofcomputerizedaxialtomography(CAT),alsocalledcomputerizedtomography(CT)forshort,isoneofthemostimportanteventsintheapplicationofimageprocessinginmedicaldiagnosis.Computerizedaxialtomographyisaprocessinwhicharingofdetectorsencirclesanobject(orpatient)andanX-raysource,concentricwiththedetectorring,rotatesabouttheobject.TheX-rayspassthroughtheobjectandarecoll