外文翻译数字图像处理方法的研究.docx
《外文翻译数字图像处理方法的研究.docx》由会员分享,可在线阅读,更多相关《外文翻译数字图像处理方法的研究.docx(14页珍藏版)》请在冰豆网上搜索。
外文翻译数字图像处理方法的研究
Theresearchofdigitalimageprocessingtechnique
1Introduction
Interestindigitalimageprocessingmethodsstemsfromtwoprincipalapplicationareas:
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,digitalimageprocessingencompassesawideandvariedfieldofapplication.
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,uptoandincludingtherecognitionofindividualobjects.Asasimpleillustrationtoclarifytheseconcepts,considertheareaofautomatedanalysisoftext.Theprocessesofacquiringanimageoftheareacontainingthetext.Preprocessingthatimages,extracting(segmenting)theindividualcharacters,describingthecharactersinaformsuitableforcomputerprocessing,andrecognizingthoseindividualcharactersareinthescopeofwhatwecalldigitalimageprocessinginthisbook.Makingsenseofthecontentofthepagemaybeviewedasbeinginthedomainofimageanalysisandevencomputervision,dependingonthelevelofcomplexityimpliedbythestatement“makingcense.”Aswillbecomeevidentshortly,digitalimageprocessing,aswehavedefinedit,isusedsuccessfullyinabroadrangofareasofexceptionalsocialandeconomicvalue.Theconceptsdevelopedinthefollowingchaptersarethefoundationforthemethodsusedinthoseapplicationareas.
1.2TheOriginsofDigitalImageProcessing
Oneofthefirstapplicationsofdigitalimageswasinthenewspaperindustry,whenpictureswerefirstsentbysubmarinecablebetweenLondonandNewYork.IntroductionoftheBartlanecablepicturetransmissionsystemintheearly1920sreducedthetimerequiredtotransportapictureacrosstheAtlanticfrommorethanaweektolessthanthreehours.Specializedprintingequipmentcodedpicturesforcabletransmissionandthenreconstructedthematthereceivingend.Figure1.1wastransmittedinthiswayandreproducedonatelegraphprinterfittedwithtypefacessimulatingahalftonepattern.
Someoftheinitialproblemsinimprovingthevisualqualityoftheseearlydigitalpictureswererelatedtotheselectionofprintingproceduresandthedistributionofintensitylevels.TheprintingmethodusedtoobtainFig.1.1wasabandonedtowardtheendof1921infavorofatechniquebasedonphotographicreproductionmadefromtapesperforatedatthetelegraphreceivingterminal.Figure1.2showsanimagesobtainedusingthismethod.TheimprovementsoverFig.1.1areevident,bothintonalqualityandinresolution.
FIGURE1.1AdigitalpictureproducedinFIGURE1.2Adigitalpicture
1921fromacodedtapebyatelegraphprintermadein1922fromatapepunched
Withspecialtypefaces(McFarlane)afterthesignalshadcrossedthe
Atlantictwice.Someerrorsare
Visible.(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,therewereaseriesofadvancesthatledtocomputerspowerfulenoughtobeusedfordigitalimageprocessing.Briefly,theseadvancesmaybesummarizedasfollow:
(1)theinventionofthetransistorbyBellLaboratoriesin1948;
(2)thedevelopmentinthe1950sand1960softhehigh-levelprogramminglanguagesCOBOL(CommonBusiness-OrientedLanguage)andFORTRAN(FormulaTranslator);
(3)theinventionoftheintegratedcircuit(IC)atTexasInstrumentsin1958;
(4)thedevelopmentofoperatingsystemintheearly1960s;
(5)thedevelopmentofthemicroprocessor(asinglechipconsistingofthecentralprocessingunit,memory,andinputandoutputcontrols)byInterintheearly1970s;
(6)introductionbyIBMofthepersonalcomputerin1981;
(7)progressiveminiaturizationofcomponents,startingwithlargescaleintegration(LI)inthelate1970s,thenverylargescaleintegration(VLSI)inthe1980s,tothepresentuseofultralargescaleintegr