THERESEARCHOFDIGITALIMAGEPROCESSINGTECHNIQUE1INTRODUCTIONINTERESTINDIGITALIMAGEPROCESSINGMETHODSSTEMSFROMTWOPRINCIPALAPPLICATIONAREASIMPROVEMENTOFPICTORIALINFORMATIONFORHUMANINTERPRETATIONANDPROCESSINGOFIMAGEDATAFORSTORAGE,TRANSMISSION,ANDREPRESENTATIONFORAUTONOMOUSMACHINEPERCEPTIONTHISCHAPTERHASSEVERALOBJECTIVES1TODEFINETHESCOPEOFTHEFIELDTHATWECALLIMAGEPROCESSING2TOGIVEAHISTORICALPERSPECTIVEOFTHEORIGINSOFTHISFIELD3TOGIVEANIDEAOFTHESTATEOFTHEARTINIMAGEPROCESSINGBYEXAMININGSOMEOFTHEPRINCIPALAREAINWHICHITISAPPLIED4TODISCUSSBRIEFLYTHEPRINCIPALAPPROACHESUSEDINDIGITALIMAGEPROCESSING5TOGIVEANOVERVIEWOFTHECOMPONENTSCONTAINEDINATYPICAL,GENERALPURPOSEIMAGEPROCESSINGSYSTEMAND6TOPROVIDEDIRECTIONTOTHEBOOKSANDOTHERLITERATUREWHEREIMAGEPROCESSINGWORKNORMALLYISREPORTER11WHATISDIGITALIMAGEPROCESSINGANIMAGEMAYBEDEFINEDASATWODIMENSIONALFUNCTION,FX,Y,WHEREXANDYARESPATIALPLANECOORDINATES,ANDTHEAMPLITUDEOFFATANYPAIROFCOORDINATESX,YISCALLEDTHEINTENSITYORGRAYLEVELOFTHEIMAGEATTHATPOINTWHENX,Y,ANDDIGITALIMAGETHEFIELDOFDIGITALIMAGEPROCESSINGREFERSTOPROCESSINGDIGITALIMAGESBYMEANSOFADIGITALCOMPUTERNOTETHATADIGITALIMAGEISCOMPOSEDOFAFINITENUMBEROFELEMENTS,EACHOFWHICHHASAPARTICULARLOCATIONANDVALUETHESEELEMENTSAREREFERREDTOASPICTUREELEMENTS,IMAGEELEMENTS,PELS,ANDPIXELSPIXELISTHETERMMOSTWIDELYUSEDTODENOTETHEELEMENTSOFADIGITALIMAGEWECONSIDERTHESEDEFINITIONSINMOREFORMALTERMSINCHAPTER2VISIONISTHEMOSTADVANCEDOFOURSENSES,SOITISNOTSURPRISINGTHATIMAGESPLAYTHESINGLEMOSTIMPORTANTROLEINHUMANPERCEPTIONHOWEVER,UNLIKEHUMANWHOARELIMITEDTOTHEVISUALBANDOFTHEELECTROMAGNETICEMSPECTRUM,IMAGINGMACHINESCOVERALMOSTTHEENTIREEMSPECTRUM,RANGINGFROMGAMMATORADIOWAVESTHEYCANOPERATEONIMAGESGENERATEDBYSOURCESTHATHUMANARENOTACCUSTOMEDTOASSOCIATINGWITHIMAGETHESEINCLUDEULTRASOUND,ELECTRONMICROSCOPY,ANDCOMPUTERGENERATEDIMAGESTHUS,DIGITALIMAGEPROCESSINGENCOMPASSESAWIDEANDVARIEDFIELDOFAPPLICATIONTHEREISNOGENERALAGREEMENTAMONGAUTHORSREGARDINGWHEREIMAGEPROCESSINGSTOPSANDOTHERRELATEDAREAS,SUCHASIMAGEANALYSISANDCOMPUTERVISION,STARTSOMETIMESADISTINCTIONISMADEBYDEFININGIMAGEPROCESSINGASADISCIPLINEINWHICHBOTHTHEINPUTANDOUTPUTOFAPROCESSAREIMAGESWEBELIEVETHISTOBEALIMITINGANDSOMEWHATARTIFICIALBOUNDARYFOREXAMPLE,UNDERTHISDEFINITION,EVENTHETRIVIALTASKOFCOMPUTINGTHEAVERAGEINTENSITYOFANIMAGEWHICHYIELDSASINGLENUMBERWOULDNOTBECONSIDEREDANIMAGEPROCESSINGOPERATIONONTHEOTHERHAND,THEREAREFIELDSSUCHASCOMPUTERVISIONWHOSEULTIMATEGOALISTOUSECOMPUTERTOEMULATEHUMANVISION,INCLUDINGLEARNINGANDBEINGABLETOMAKEINFERENCESANDTAKEACTIONSBASEDONVISUALINPUTSTHISAREAITSELFISABRANCHOFARTIFICIALINTELLIGENCEAIWHOSEOBJECTIVEISTOEMULATEHUMANINTELLIGENCETHISFIELDOFAIISINITSEARLIESTSTAGESOFINFANCYINTERMSOFDEVELOPMENT,WITHPROGRESSHAVINGBEENMUCHSLOWERTHANORIGINALLYANTICIPATEDTHEAREAOFIMAGEANALYSISALSOCALLEDIMAGEUNDERSTANDINGISINBETWEENIMAGEPROCESSINGANDCOMPUTERVISIONTHEREARENOCLEARCUTBOUNDARIESINTHECONTINUUMFROMIMAGEPROCESSINGATONEENDTOCOMPUTERVISIONATTHEOTHERHOWEVER,ONEUSEFULPARADIGMISTOCONSIDERTHREETYPESOFCOMPUTERIZEDPROCESSESISTHISCONTINUUMLOW,MID,ANDHIGHEVERPROCESSESLOWLEVELPROCESSESINVOLVEPRIMITIVEOPERATIONSUCHASIMAGEPREPROCESSINGTOREDUCENOISE,CONTRASTENHANCEMENT,ANDIMAGESHARPENINGALOWLEVELPROCESSISCHARACTERIZEDBYTHEFACTTHATBOTHITSINPUTANDOUTPUTAREIMAGESMIDLEVELPROCESSINGONIMAGESINVOLVESTASKSSUCHASSEGMENTATIONPARTITIONINGANIMAGEINTOREGIONSOROBJECTS,DESCRIPTIONOFTHOSEOBJECTSTOREDUCETHEMTOAFORMSUITABLEFORCOMPUTERPROCESSING,ANDCLASSIFICATIONRECOGNITIONOFINDIVIDUALOBJECTAMIDLEVELPROCESSISCHARACTERIZEDBYTHEFACTTHATITSINPUTSGENERALLYAREIMAGES,BUTITSOUTPUTISATTRIBUTESEXTRACTEDFROMTHOSEIMAGESEG,EDGESCONTOURS,ANDTHEIDENTITYOFINDIVIDUALOBJECTFINALLY,HIGHERLEVELPROCESSINGINVOLVES“MAKINGSENSE”O(jiān)FANENSEMBLEOFRECOGNIZEDOBJECTS,ASINIMAGEANALYSIS,AND,ATTHEFARENDOFTHECONTINUUM,PERFORMINGTHECOGNITIVEFUNCTIONNORMALLYASSOCIATEDWITHVISIONBASEDONTHEPRECEDINGCOMMENTS,WESEETHATALOGICALPLACEOFOVERLAPBETWEENIMAGEPROCESSINGANDIMAGEANALYSISISTHEAREAOFRECOGNITIONOFINDIVIDUALREGIONSOROBJECTSINANIMAGETHUS,WHATWECALLINTHISBOOKDIGITALIMAGEPROCESSINGENCOMPASSESPROCESSESWHOSEINPUTSANDOUTPUTSAREIMAGESAND,INADDITION,ENCOMPASSESPROCESSESTHATEXTRACTATTRIBUTESFROMIMAGES,UPTOANDINCLUDINGTHERECOGNITIONOFINDIVIDUALOBJECTSASASIMPLEILLUSTRATIONTOCLARIFYTHESECONCEPTS,CONSIDERTHEAREAOFAUTOMATEDANALYSISOFTEXTTHEPROCESSESOFACQUIRINGANIMAGEOFTHEAREACONTAININGTHETEXTPREPROCESSINGTHATIMAGES,EXTRACTINGSEGMENTINGTHEINDIVIDUALCHARACTERS,DESCRIBINGTHECHARACTERSINAFORMSUITABLEFORCOMPUTERPROCESSING,ANDRECOGNIZINGTHOSEINDIVIDUALCHARACTERSAREINTHESCOPEOFWHATWECALLDIGITALIMAGEPROCESSINGINTHISBOOKMAKINGSENSEOFTHECONTENTOFTHEPAGEMAYBEVIEWEDASBEINGINTHEDOMAINOFIMAGEANALYSISANDEVENCOMPUTERVISION,DEPENDINGONTHELEVELOFCOMPLEXITYIMPLIEDBYTHESTATEMENT“MAKINGCENSE”ASWILLBECOMEEVIDENTSHORTLY,DIGITALIMAGEPROCESSING,ASWEHAVEDEFINEDIT,ISUSEDSUCCESSFULLYINABROADRANGOFAREASOFEXCEPTIONALSOCIALANDECONOMICVALUETHECONCEPTSDEVELOPEDINTHEFOLLOWINGCHAPTERSARETHEFOUNDATIONFORTHEMETHODSUSEDINTHOSEAPPLICATIONAREAS12THEORIGINSOFDIGITALIMAGEPROCESSINGONEOFTHEFIRSTAPPLICATIONSOFDIGITALIMAGESWASINTHENEWSPAPERINDUSTRY,WHENPICTURESWEREFIRSTSENTBYSUBMARINECABLEBETWEENLONDONANDNEWYORKINTRODUCTIONOFTHEBARTLANECABLEPICTURETRANSMISSIONSYSTEMINTHEEARLY1920SREDUCEDTHETIMEREQUIREDTOTRANSPORTAPICTUREACROSSTHEATLANTICFROMMORETHANAWEEKTOLESSTHANTHREEHOURSSPECIALIZEDPRINTINGEQUIPMENTCODEDPICTURESFORCABLETRANSMISSIONANDTHENRECONSTRUCTEDTHEMATTHERECEIVINGENDFIGURE11WASTRANSMITTEDINTHISWAYANDREPRODUCEDONATELEGRAPHPRINTERFITTEDWITHTYPEFACESSIMULATINGAHALFTONEPATTERNSOMEOFTHEINITIALPROBLEMSINIMPROVINGTHEVISUALQUALITYOFTHESEEARLYDIGITALPICTURESWERERELATEDTOTHESELECTIONOFPRINTINGPROCEDURESANDTHEDISTRIBUTIONOFINTENSITYLEVELSTHEPRINTINGMETHODUSEDTOOBTAINFIG11WASABANDONEDTOWARDTHEENDOF1921INFAVOROFATECHNIQUEBASEDONPHOTOGRAPHICREPRODUCTIONMADEFROMTAPESPERFORATEDATTHETELEGRAPHRECEIVINGTERMINALFIGURE12SHOWSANIMAGESOBTAINEDUSINGTHISMETHODTHEIMPROVEMENTSOVERFIG11AREEVIDENT,BOTHINTONALQUALITYANDINRESOLUTIONFIGURE11ADIGITALPICTUREPRODUCEDINFIGURE12ADIGITALPICTURE1921FROMACODEDTAPEBYATELEGRAPHPRINTERMADEIN1922FROMATAPEPUNCHEDWITHSPECIALTYPEFACESMCFARLANEAFTERTHESIGNALSHADCROSSEDTHEATLANTICTWICESOMEERRORSAREVISIBLEMCFARLANETHEEARLYBARTLANESYSTEMSWERECAPABLEOFCODINGIMAGESINFIVEDISTINCTLEVELOFGRAYTHISCAPABILITYWASINCREASEDTO15LEVELSIN1929FIGURE13ISTYPICALOFTHEIMAGESTHATCOULDBEOBTAINEDUSINGTHE15TONEEQUIPMENTDURINGTHISPERIOD,INTRODUCTIONOFASYSTEMFORDEVELOPINGAFILMPLATEVIALIGHTBEAMSTHATWEREMODULATEDBYTHECODEDPICTURETAPEIMPROVEDTHEREPRODUCTIONPROCESSCONSIDERABLYALTHOUGHTHEEXAMPLESJUSTCITEDINVOLVEDIGITALIMAGES,THEYARENOTCONSIDEREDDIGITALIMAGEPROCESSINGRESULTSINTHECONTEXTOFOURDEFINITIONBECAUSECOMPUTERWERENOTINVOLVEDINTHEIRCREATIONTHUS,THEHISTORYOFDIGITALPROCESSINGISINTIMATELYTIEDTOTHEDEVELOPMENTOFTHEDIGITALCOMPUTERINFACTDIGITALIMAGESREQUIRESOMUCHSTORAGEANDCOMPUTATIONALPOWERTHATPROGRESSINTHEFIELDOFDIGITALIMAGEPROCESSINGHASBEENDEPENDENTONTHEDEVELOPMENTOFDIGITALCOMPUTERSOFSUPPORTINGTECHNOLOGIESTHATINCLUDEDATASTORAGE,DISPLAY,ANDTRANSMISSIONTHEIDEAOFACOMPUTERGOESBACKTOTHEINVENTIONOFTHEABACUSINASIAMINOR,MORETHAN5000YEARSAGOMORERECENTLY,THEREWEREDEVELOPMENTSINTHEPASTTWOCENTURIESTHATARETHEFOUNDATIONOFWHATWECALLCOMPUTERTODAYHOWEVER,THEBASISFORWHATWECALLAMODERNDIGITALCOMPUTERDATESBACKTOONLYTHE1940SWITHTHEINTRODUCTIONBYJOHNVONNEUMANNOFTWOKEYCONCEPTS1AMEMORYTOHOLDASTOREDPROGRAMANDDATA,AND2CONDITIONALBRANCHINGTHERETWOIDEASARETHEFOUNDATIONOFACENTRALPROCESSINGUNITCPU,WHICHISATTHEHEARTOFCOMPUTERTODAYSTARTINGWITHVONNEUMANN,THEREWEREASERIESOFADVANCESTHATLEDTOCOMPUTERSPOWERFULENOUGHTOBEUSEDFORDIGITALIMAGEPROCESSINGBRIEFLY,THESEADVANCESMAYBESUMMARIZEDASFOLLOW1THEINVENTIONOFTHETRANSISTORBYBELLLABORATORIESIN19482THEDEVELOPMENTINTHE1950SAND1960SOFTHEHIGHLEVELPROGRAMMINGLANGUAGESCOBOLCOMMONBUSINESSORIENTEDLANGUAGEANDFORTRANFORMULATRANSLATOR3THEINVENTIONOFTHEINTEGRATEDCIRCUITICATTEXASINSTRUMENTSIN19584THEDEVELOPMENTOFOPERATINGSYSTEMINTHEEARLY1960S5THEDEVELOPMENTOFTHEMICROPROCESSORASINGLECHIPCONSISTINGOFTHECENTRALPROCESSINGUNIT,MEMORY,ANDINPUTANDOUTPUTCONTROLSBYINTERINTHEEARLY1970S6INTRODUCTIONBYIBMOFTHEPERSONALCOMPUTERIN19817PROGRESSIVEMINIATURIZATIONOFCOMPONENTS,STARTINGWITHLARGESCALEINTEGRATIONLIINTHELATE1970S,THENVERYLARGESCALEINTEGRATIONVLSIINTHE1980S,TOTHEPRESENTUSEOFULTRALARGESCALEINTEGRATIONULSIFIGURE13IN1929FROMLONDONTOCENERALEPERSHINGTHATNEWYORKDELIVERSWITH15LEVELTONEEQUIPMENTSTHROUGHCABLEWITHFOCHDONOTTHEPHOTOGRAPHBYDECORATIONCONCURRENTWITHTHESEADVANCESWEREDEVELOPMENTINTHEAREASOFMASSSTORAGEANDDISPLAYSYSTEMS,BOTHOFWHICHAREFUNDAMENTALREQUIREMENTSFORDIGITALIMAGEPROCESSINGTHEFIRSTCOMPUTERSPOWERFULENOUGHTOCARRYOUTMEANINGFULIMAGEPROCESSINGTASKSAPPEAREDINTHEEARLY1960STHEBIRTHOFWHATWECALLDIGITALIMAGEPROCESSINGTODAYCANBETRACEDTOTHEAVAILABILITYOFTHOSEMACHINESANDTHEONSETOFTHEAPACEPROGRAMDURINGTHATPERIODITTOOKTHECOMBINATIONOFTHOSETWODEVELOPMENTSTOBRINGINTOFOCUSTHEPOTENTIALOFDIGITALIMAGEPROCESSINGCONCEPTSWORKONUSINGCOMPUTERTECHNIQUESFORIMPROVINGIMAGESFROMASPACEPROBEBEGANATTHEJETPROPULSIONLABORATORYPASADENA,CALIFORNIAIN1964WHENPICTURESOFTHEMOONTRANSMITTEDBYRANGER7WEREPROCESSEDBYACOMPUTERTOCORRECTVARIOUSTYPESOFIMAGEDISTORTIONINHERENTINTHEONBOARDTELEVISIONCAMERAFIGURE14SHOWSTHEFIRSTIMAGEOFTHEMOONTAKENBYRANGER7ONJULY31,1964AT909AMEASTERNDAYLIGHTTIMEEDT,ABOUT17MINUTESBEFOREIMPACTINGTHELUNARSURFACETHEMARKERS,CALLEDRESEAUMARK,AREUSEDFORGEOMETRICCORRECTIONS,ASDISCUSSEDINCHAPTER5THISALSOISTHEFIRSTIMAGEOFTHEMOONTAKENBYAUSSPACECRAFTTHEIMAGINGLESSONSLEARNEDWITHRANGER7SERVEDASTHEBASISFORIMPROVEDMETHODSUSEDTOENHANCEANDRESTOREIMAGESFROMTHESURVEYORMISSIONSTOTHEMOON,THEMARINERSERIESOFFLYBYMISSIONTOMARS,THEAPOLLOMANNEDFLIGHTSTOTHEMOON,ANDOTHERSINPARALLELWITHSPACEAPPLICATION,DIGITALIMAGEPROCESSINGTECHNIQUESBEGANINTHELATE1960SANDEARLY1970STOBEUSEDINMEDICALIMAGING,REMOTEEARTHRESOURCESOBSERVATIONS,ANDASTRONOMYTHEINVENTIONINTHEEARLY1970SOFCOMPUTERIZEDAXIALTOMOGRAPHYCAT,ALSOCALLEDCOMPUTERIZEDTOMOGRAPHYCTFORSHORT,ISONEOFTHEMOSTIMPORTANTEVENTSINTHEAPPLICATIONOFIMAGEPROCESSINGINMEDICALDIAGNOSISCOMPUTERIZEDAXIALTOMOGRAPHYISAPROCESSINWHICHARINGOFDETECTORSENCIRCLESANOBJECTORPATIENTANDANXRAYSOURCE,CONCENTRICWITHTHEDETECTORRING,ROTATESABOUTTHEOBJECTTHEXRAYSPASSTHROUGHTHEOBJECTANDARECOLLECTEDATTHEOPPOSITEENDBYTHECORRESPONDINGDETECTORSINTHERINGASTHESOURCEROTATES,THISPROCEDUREISREPEATEDTOMOGRAPHYCONSISTSOFALGORITHMSTHATUSETHESENSEDDATATOCONSTRUCTANIMAGETHATREPRESENTSA“SLICE”THROUGHTHEOBJECTMOTIONOFTHEOBJECTINADIRECTIONPERPENDICULARTOTHERINGOFDETECTORSPRODUCESASETOFSUCHSLICES,WHICHCONSTITUTEATHREEDIMENSIONAL3DRENDITIONOFTHEINSIDEOFTHEOBJECTTOMOGRAPHYWASINVENTEDINDEPENDENTLYBYSIRGODFREYNHOUNSFIELDANDPROFESSORALLANMCORMACK,WHOSHAREDTHEXRAYSWEREDISCOVEREDIN1895BYWILHELMCONRADROENTGEN,FORWHICHHERECEIVEDTHE1901NOBELPRIZEFORPHYSICSTHESETWOINVENTIONS,NEARLY100YEARSAPART,LEDTOSOMEOFTHEMOSTACTIVEAPPLICATIONAREASOFIMAGEPROCESSINGTODAYFIGURE14THEFIRSTPICTUREOFTHEMOONBYAUSSPACECRAFTRANGER7TOOKTHISIMAGEONJULY31,1964AT909AMEDT,ABOUT17MINUTESBEFOREIMPACTINGTHELUNARSURFACECOURTESYOFNASA數(shù)字圖像處理方法的研究1緒論數(shù)字圖像處理方法的研究源于兩個主要應(yīng)用領(lǐng)域其一是為了便于人們分析而對圖像信息進(jìn)行改進(jìn)；其二是為了使機(jī)器自動理解而對圖像數(shù)據(jù)進(jìn)行存儲、傳輸及顯示。11數(shù)字圖像處理的概念一幅圖像可定義為一個二維函數(shù)FX,Y，這里X和Y是空間坐標(biāo)，而在任何一對空間坐標(biāo)FX,Y上的幅值F稱為該點(diǎn)圖像的強(qiáng)度或灰度。當(dāng)X，Y和幅值F為有限的、離散的數(shù)值時，稱該點(diǎn)是由有限的元素組成的，沒一個元素都有一個特定的位置和幅值，這些元素稱為圖像元素、畫面元素或象素。象素是廣泛用于表示數(shù)字圖像元素的詞匯。在第二章，將用更正式的術(shù)語研究這些定義。視覺是人類最高級的感知器官，所以，毫無疑問圖像在人類感知中扮演著最重要的角色。然而，人類感知只限于電磁波譜的視覺波段，成像機(jī)器則可覆蓋幾乎全部電磁波譜，從伽馬射線到無線電波。它們可以對非人類習(xí)慣的那些圖像源進(jìn)行加工，這些圖像源包括超聲波、電子顯微鏡及計算機(jī)產(chǎn)生的圖像。因此，數(shù)字圖像處理涉及各種各樣的應(yīng)用領(lǐng)域。圖像處理涉及的范疇或其他相關(guān)領(lǐng)域例如，圖像分析和計算機(jī)視覺的界定在初創(chuàng)人之間并沒有一致的看法。有時用處理的輸人和輸出內(nèi)容都是圖像這一特點(diǎn)來界定圖像處理的范圍。我們認(rèn)為這一定義僅是人為界定和限制。例如，在這個定義下，甚至最普通的計算一幅圖像灰度平均值的工作都不能算做是圖像處理。另一方面，有些領(lǐng)域如計算機(jī)視覺研究的最高目標(biāo)是用計算機(jī)去模擬人類視覺，包括理解和推理并根據(jù)視覺輸人采取行動等。這一領(lǐng)域本身是人工智能的分支，其目的是模仿人類智能。人工智能領(lǐng)域處在其發(fā)展過程中的初期階段，它的發(fā)展比預(yù)期的要慢得多，圖像分析也稱為圖像理解領(lǐng)域則處在圖像處理和計算機(jī)視覺兩個學(xué)科之間。從圖像處理到計算機(jī)視覺這個連續(xù)的統(tǒng)一體內(nèi)并沒有明確的界線。然而，在這個連續(xù)的統(tǒng)一體中可以考慮三種典型的計算處理即低級、中級和高級處理來區(qū)分其中的各個學(xué)科。低級處理涉及初級操作，如降低噪聲的圖像預(yù)處理，對比度增強(qiáng)和圖像尖銳化。低級處理是以輸人、輸出都是圖像為特點(diǎn)的處理。中級處理涉及分割把圖像分為不同區(qū)域或目標(biāo)物以及縮減對目標(biāo)物的描述，以使其更適合計算機(jī)處理及對不同日標(biāo)的分類識別。中級圖像處理是以輸人為圖像，但輸出是從這些圖像中提取的特征如邊緣、輪廓及不同物體的標(biāo)識等為特點(diǎn)的。最后，高級處理涉及在圖像分析中被識別物體的總體理解，以及執(zhí)行與視覺相關(guān)的識別函數(shù)處在連續(xù)統(tǒng)一體邊緣等。根據(jù)上述討論，我們看到，圖像處理和圖像分析兩個領(lǐng)域合乎邏輯的重疊區(qū)域是圖像中特定區(qū)域或物體的識別這一領(lǐng)域。這樣，在本書中，我們界定數(shù)字圖像處理包括輸人和輸出均是圖像的處理，同時也包括從圖像中提取特征及識別特定物體的處理。舉一個簡單的文本自動分析方面的例子來具體說明這一概念。在自動分析文本時首先獲取一幅包含文本的圖像，對該圖像進(jìn)行預(yù)處理，提取分割字符，然后以適合計算機(jī)處理的形式描述這些字符，最后識別這些字符，而所有這些操作都在本書界定的數(shù)字圖像處理的范圍內(nèi)。理解一頁的內(nèi)容可能要根據(jù)理解的復(fù)雜度從圖像分析或計算機(jī)視覺領(lǐng)域考慮問題。這樣，本書定義的數(shù)字圖像處理的概念將在有特殊社會和經(jīng)濟(jì)價值的領(lǐng)域內(nèi)通用。在以下各章展開的概念是那些應(yīng)用領(lǐng)域所用方法的基礎(chǔ)。12數(shù)字圖像處理的起源數(shù)字圖像處理最早的應(yīng)用之一是在報紙業(yè)，當(dāng)時，圖像第一次通過海底電纜從倫敦傳往紐約。早在20世紀(jì)20年代曾引入BTUTLANE電纜圖片傳輸系統(tǒng)，把橫跨大西洋傳送一幅圖片所需的時間從一個多星期減少到3個小時。為了用電纜傳輸圖片，首先要進(jìn)行編碼，然后在接收端用特殊的打印設(shè)備重構(gòu)該圖片。圖11就是用這種方法傳送并利用電報打印機(jī)通過字符模擬中間色調(diào)還原出來的圖像。這些早期數(shù)字圖像視覺質(zhì)量的改進(jìn)工作，涉及到打印過程的選擇和亮度等級的分布等問題。用于得到圖11的打印方法到1921年底就被徹底淘汰了，轉(zhuǎn)而支持一種基于光學(xué)還原的技術(shù)，該技術(shù)在電報接收端用穿孔紙帶打出圖片。圖12就是用這種方法得到的圖像，對比圖11，它在色調(diào)質(zhì)量和分辨率方面的改進(jìn)都很明顯。圖111421年由電報打印機(jī)采用特殊字圖121922年在信號兩次穿越大西洋后，符在編碼紙帶中產(chǎn)生的數(shù)字圖像從穿孔紙帶得到的數(shù)字圖像，可以MCFALSNE看出某些差錯MCFALSNE早期的BARTLANE系統(tǒng)可以用5個灰度等級對圖像編碼，到1929年已增加到15個等級。圖13所示的這種典型類型的圖像就是用15級色調(diào)設(shè)備得到的。在這一時期，由于引入了一種用編碼圖像紙帶去調(diào)制光束而使底片感光的系統(tǒng)，明顯地改善了復(fù)原過程。剛才引用的數(shù)字圖像的例子并沒有考慮數(shù)字圖像處理的結(jié)果，這主要是因為沒有涉及到計算機(jī)。因此，數(shù)字圖像處理的歷史與數(shù)字計算機(jī)的發(fā)展密切相關(guān)。事實上，數(shù)字圖像要求非常大的存儲和計算能力，因此數(shù)字圖像處理領(lǐng)域的發(fā)展必須依靠數(shù)字計算機(jī)及數(shù)據(jù)存儲、顯示和傳輸?shù)认嚓P(guān)技術(shù)的發(fā)展。計算機(jī)的概念可追溯到5000多年前中國算盤的發(fā)明。近兩個世紀(jì)以來的一些發(fā)展也奠定了計算機(jī)的基礎(chǔ)。然而，現(xiàn)代計算機(jī)的基礎(chǔ)還要回溯到20世紀(jì)40年代由約翰馮諾依曼提出的兩個重要概念L保存程序和數(shù)據(jù)的存儲器2條件分支。這兩個概念是中央處理單元CPU的基礎(chǔ)。今天，它是計算機(jī)的心臟。從馮諾依曼開始，引發(fā)了一系列重要技術(shù)進(jìn)步，使得計算機(jī)以強(qiáng)大的功能用于數(shù)字圖像處理領(lǐng)域。簡單說，這些進(jìn)步可歸納為如下幾點(diǎn)11948年貝爾實驗室發(fā)明了晶體三極管；220世紀(jì)50年代到20世紀(jì)60年代高級編程語言如COBOL和FORTRAN的開發(fā)；31958年得州儀器公司發(fā)明了集成電路IC420世紀(jì)60年代早期操作系統(tǒng)的發(fā)展520世紀(jì)70年代INTEL公司開發(fā)了微處理器由中央處理單元、存儲器和輸入輸出控制組成的單一芯片61981年IBM公司推出了個人計算機(jī)720世紀(jì)70年代出現(xiàn)的大規(guī)模集成電路LI所引發(fā)的元件微小化革命，20世紀(jì)80年代出現(xiàn)了YLSI超大規(guī)模集成電路，現(xiàn)在已出現(xiàn)了ULSI。圖13在1929年從倫敦到紐約用15級色調(diào)設(shè)備通過電纜傳送的CENERALEPERSHING和FOCH的未經(jīng)修飾的照片伴隨著這些技術(shù)進(jìn)步，大規(guī)模的存儲和顯示系統(tǒng)也隨之發(fā)展起來。這兩者均是數(shù)字圖像處理的基礎(chǔ)。第一臺可以執(zhí)行有意義的圖像處理任務(wù)的大型計算機(jī)出現(xiàn)在20世紀(jì)60年代早期。數(shù)字圖像處理技術(shù)的誕生可追溯至這一時期這些機(jī)器的使用和空間項目的開發(fā)，這兩大發(fā)展把人們的注意力集中到數(shù)字圖像處理的潛能上。利用計算機(jī)技術(shù)改善空間探測器發(fā)回的圖像的工作，始于1964年美國加利福尼亞的噴氣推進(jìn)實驗室。當(dāng)時由“旅行者7號”衛(wèi)星傳送的月球圖像由一臺計算機(jī)進(jìn)行了處理，以校正航天器上電視攝像機(jī)中各種類型的圖像畸變。圖14顯示了由“旅行者7號”于1954年7月31日上午東部白天時間9點(diǎn)09分在光線影響月球表面前約17分鐘時攝取的第一張月球圖像痕跡稱為網(wǎng)狀痕跡用于幾何校正，在第5章將討論該間題，這也是美國航天器取得的第一幅月球圖像?！奥眯姓?號”傳送的圖像可作為改善的增強(qiáng)和復(fù)原圖像例如來自“探索者”登月一飛行、“水手號”系列空間探淵器及阿波羅載人登月飛行的圖像方法的基礎(chǔ)。進(jìn)行空間應(yīng)用的同時，數(shù)字圖像處理技術(shù)在20世紀(jì)60年代末和20世紀(jì)70年代初開始用于醫(yī)學(xué)圖像、地球遙感監(jiān)測和天文學(xué)等領(lǐng)域。早在20世紀(jì)70年代發(fā)明的計算機(jī)軸向斷層術(shù)CAT簡稱計算機(jī)斷層CT是圖像處理在醫(yī)學(xué)診斷領(lǐng)域最重要的應(yīng)用之一。計算機(jī)軸向斷層術(shù)是一種處理方法，在這種處理中，一個檢測器環(huán)圍繞著一個物體或病人，并且一個X射線源與檢測器環(huán)同心繞著物體旋轉(zhuǎn)。X射線穿過物體并由位于對面環(huán)中的相應(yīng)檢測器收集起來。當(dāng)X射線源旋轉(zhuǎn)時，重復(fù)這一過程。斷層技術(shù)由一些算法組成，該算法用感知的數(shù)據(jù)去重建通過物體的“切片”圖像。當(dāng)物體沿垂直于檢測器的方向運(yùn)動時就產(chǎn)生一系列這樣的“切片”，這些切片組成了物體內(nèi)部的再現(xiàn)圖像。斷層技術(shù)是由GODFREYNHOUNSFIELD先生和ALLANMCORMACK教授發(fā)明的，他們共同獲得了1979年諾貝爾醫(yī)學(xué)獎。X射線是在1895年由威廉康拉德倫琴發(fā)現(xiàn)的，由于這一發(fā)現(xiàn)，他獲得了I901年諾貝爾物理學(xué)獎。這兩項發(fā)明相差近100年。它們在今天引領(lǐng)著圖像處理某些最活躍的應(yīng)用領(lǐng)域。圖14美國航天器傳送的第一張月球照片，“旅行者7號”衛(wèi)星1964年7月31日9點(diǎn)09分東部白天時間在光線影響月球表面前17分鐘時攝取的圖像10/301712C620軸撥桿的工藝規(guī)程及鉆216孔的鉆床夾具設(shè)計09/211339CA6140車床撥叉零件的機(jī)械加工工藝規(guī)程及夾具設(shè)計83100308/301537CPU風(fēng)扇后蓋的注塑模具設(shè)計09/201619GDC956160工業(yè)對輥成型機(jī)設(shè)計08/301545LS型螺旋輸送機(jī)的設(shè)計10/072343LS型螺旋輸送機(jī)設(shè)計09/201623P90B型耙斗式裝載機(jī)設(shè)計09/082017PE10自行車無級變速器設(shè)計10/070923話機(jī)機(jī)座下殼模具的設(shè)計與制造09/082020T108噸自卸車拐軸的斷裂原因分析及優(yōu)化設(shè)計09/211339XY型數(shù)控銑床工作臺的設(shè)計09/082025YD5141SYZ后壓縮式垃圾車的上裝箱體設(shè)計10/070920ZH1115W柴油機(jī)氣缸體三面粗鏜組合機(jī)床總體及左主軸箱設(shè)計09/211534ZXT06型多臂機(jī)凸輪軸加工工藝及工裝設(shè)計10/301604三孔連桿零件的工藝規(guī)程及鉆35H6孔的夾具設(shè)計08/301757三層貨運(yùn)電梯曳引機(jī)及傳動系統(tǒng)設(shè)計10/291408