圖像與處Imageand第6講：圖像增強(qiáng)-空域?yàn)V人工智能與機(jī)器InstituteofArtificialInligenceand圖像 處Image IntensityImageLogGammaPiecewise-linearHistogramHistogramHistogramLocalhistogramSpatialSpatialcorrelationandSmoothinglinearfilters&SpatialNonlinear圖像 處Image SharpeningSpatialLaplacianUnsharpMaskingandHighboostUsingFirst-OrderDerivativesforNonlinear —The圖像 處Image SharpeningSpatialFilters:-highlighttransitionsin-Electronicprinting,medialimaging,inspection,autonomous-sharpen–Thefirst-orderderivativeofaone-dimensionalfunctionistheff(x1)fThesecond-orderderivativeoff(x)asthe2f

f(x1)f(x1)2fFirstFirst-zero:-nonzero:oframpand-nonzero:secondsecond-zero:-nonzero:oframpand-zero:constant頻圖像與頻ff(x1)fblack white圖像 處Image First:detectthepresenceanedgeataZerocrossing:edgeSecond:Zerocrossing:edgeisondarkor圖像 處Image 圖像 處Image CharacteristicsofFirstandSecondFirst-orderderivativesgenerallyproduceedgesinSecond-orderderivativeshaveastrongerresponsetofinedetail,suchasthinlines,isolatedpoints,andSecond-orderderivativesproduceadouble-responseatrampandsteptransitioninThesignofthesecondderivativecanbeusedtodeterminewhetheratransitionintoanedgeisfromlighttodarkordarktolight圖像 處Image SharpeningSpatialFilters:LaplaceThesecond-orderisotropicderivativeoperatorisRotationLaplacianforafunction(image)Rotation2f

2f

2 2x22f

f(x1,y)f(x1,y)2f(x,f(x,y1)f(x,y1)2f(x,2ff(x1,y)f(x1,y)f(x,y1)f(x,y-4f(x,圖像 處Image SharpeningSpatialFilters:Laplace圖像 處Image SharpeningSpatialFilters:LaplaceImagesharpeninginthewayofusingtheg(x,y)f(x,y)c2f(x,f(x,y)isinputg(x,y)issharpenendc-1if2f(x,y)correspondingtoFig.3.37(a)or(b)andc1ifeitheroftheothertwofiltersisused.HighlightHighlightintensityDeemphasizeregionswithslowlyvarying處圖像處Image NegativevaluesClippedto0fordisplay圖像 處Image Laplace3*3,3*3,15*15,35*35,75*75,圖像 處Image UnsharpMaskingandHighboostUnsharpSharpenimagesconsistsofsubtractinganunsharpversionofanimagefromtheoriginale.g.,printingandpublishingBlurtheoriginalSubtracttheblurredimagefromtheAddthemasktothe圖像 處Image UnsharpMaskingandHighboostLetf(x,y)denotetheblurredimage,unsharpmaskinggmask(x,y)f(x,y)f(x,Thenaddaweightedportionofthemaskbacktotheg(x,y)f(x,y)k*gmask(x, kwhenk1,theprocessisreferredtoashighboost圖像 處Image UnsharpMasking:SecondSecond圖像 處UIedercegandHighboostFiltering:圖像 處Image 圖像 處Image ImageSharpeningbasedonFirst-OrderForfunctionf(x,y),thegradientoff atcoordinates(x,y)isdefinedasfgx xfgrad(f)gf y

gx22yThemagnitudeofvectorf,denotedasM(x,y)M(xgx22yGradient圖像 處Image ImageSharpeningbasedonFirst-Ordergx22yThemagnitudeofvectorf,denotedasM(x,y)M(xgx22yM(x,y)|gx||gy M(x,y) 圖像 處Image ImageSharpeningbasedonFirst-OrderM(x,y)|z9z5||z8z6SobelM(x,y)|(z72z8z9)|

)

2z4z7)|圖像 處Image 圖像 處Image ImageSharpeningbasedonFirst-OrderSumSumofmaskis0,NochangeforConstantintensity圖像 處Image HighlightHighlightEliminateslowlychangingbackground圖像 處Image Originaland圖像 處SobelandImageSobelandImage 圖像 處Image PrewittandImage 圖像 處Image Low-passSpatialHigh-passSpatialFilteringOriginalimages圖像 處Image High-passSpatialImage 圖像 處Image HighBoost理s圖像 理sExe

Enhancetheimagebysharpeningitandbybringingoutmoreoftheskeletaldetail

理s圖像 理sImage

Enhancetheimagebysharpeningitandbybringingoutmoreoftheskeletaldetail

圖像 處Image Whatcanwedowith圖像 處Igedeorcg圖像 處-=Image 圖像 處Image CrowdSynthesis(withPoojaDobackgroundsubtractionineachFindandrecordForsynthesis,randomlysampletheblobs,takingcarenottooverlapthem圖像 處Image Background

onblue圖像 處Image BlueImage 圖像 處Image Morphing=ObjectTheaimistofind“anaverage”betweentwoNotanaverageoftwoimagesof…butanimageoftheaverageHowcanwemakeasmoothtransitioninDoa“weightedaverage”overtimeHowdoweknowwhattheaverageobjectlooksWehaven’taButwecanoftenfakesomethingUsuallyrequireduser/artist圖像 處Igedo圖像 處Alignfirst,thencross-Alignmentusingglobalwarp–picturestill圖像 處Image PainterAnselAnselfactualcommunicationofideas.ItisaImageImage處圖像 處That’sImageThat’s圖像 處Image IntensityImageLogGammaPiecewise-linearHistogramHistogramHistogramLocalhist