




版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請(qǐng)進(jìn)行舉報(bào)或認(rèn)領(lǐng)
文檔簡介
第四章數(shù)字圖像處理2.圖像增強(qiáng)ImageEnhancement:TheimagewillbesignificantlyimprovedifoneormoreofthefunctionscalledEnhancementareapplied.圖像增強(qiáng):突出相關(guān)的專題信息,提高圖像的視覺應(yīng)用,使分析者更容易識(shí)別圖像內(nèi)容,從圖像中提取更有用的定量化信息; 輻射增強(qiáng)、空間增強(qiáng)以及光譜增強(qiáng)(變換)遙感影像灰度的頻率分布輻射增強(qiáng)遙感影像灰度的頻率分布CommonSymmetricandSkewedDistributionsinRSDataCommonly,thedistributionofDNs(graylevels)canbeunimodalandmaybeGaussian,althoughskewingisusual.Multimodaldistributionsresultifascenecontainstwoormoredominantclasseswithdistinctlydifferent(oftennarrow)rangesofreflectance.Min-Max
ContrastStretch+1StandardDeviationContrastStretchLinearexpansionofDN'sintothefullscale(0-255)isacommonoption.對(duì)比度拉伸
Thestartingpointistopointoutthatradiancesmeasuredbythesensorsvaryinintensity.Thevaluesarerestatedas(DNs)thatconsistofequalincrementsoverarange.ThesimplestmanipulationoftheseDNsistoincreaseordecreasetherangeofDNspresentandassignthisnewrangethegraylevelsavailablewithintherangelimit.LinearContrastEnhancement:Minimum-MaximumContrastStretchwhere:-BVinistheoriginalinputbrightnessvalue-quantkistherangeofthebrightnessvaluesthatcanbedisplayedontheCRT(e.g.,255),minkistheminimumvalueintheimage,
maxkisthemaximumvalueintheimage,and
BVoutistheoutputbrightnessvalueContrastStretchof
Landsat
TMBand4DataOriginalMinimum-maximum+1standarddeviationAllotheroriginalbrightnessvaluesbetween5and104arelinearlydistributedbetween0and255.LinearContrastEnhancement:Minimum-MaximumContrastStretchOriginalLinearNon-linearContrastStretchingThesearemostlynonlinearfunctionsthataffecttheprecisedistributionofdensities(onfilm)orgraylevels(inmonitorimage)indifferentways,sothatsomeexperimentationmayberequiredtooptimizeresults.Commonlyusedspecialstretchesinclude:1)PiecewiseLinear,2)LinearwithSaturation3)Logarithmic,4)Exponential5)ProbabilityDistributionFunction,and…LogarithmicandInverseLogPiecewisePiecewiseLinearContrastStretchingContrastStretchingofPredawnThermalInfraredDataofthetheSavannahRiverOriginalMinimum-maximum+1standarddeviation
LinearwithSaturationHistogramEqualization直方圖均衡
evaluatestheindividualbrightnessvaluesinabandofimageryandassignsapproximatelyanequalnumberofpixelstoeachoftheuser-specifiedoutputgray-scaleclasses(e.g.,32,64,and256).
appliesthegreatestcontrastenhancementtothemostpopulatedrangeofbrightnessvaluesintheimage.
reducesthecontrastintheverylightordarkpartsoftheimageassociatedwiththetailsofanormallydistributedhistogram.Statisticsfora64x64HypotheticalImagewithBrightnessValuesfrom0to74096totalHistogramEqualization5TransformationFunction,
ki
foreachindividualbrightnessvalueForeachbrightnessvaluelevelBViinthequantkrangeof0to7oftheoriginalhistogram,anewcumulativefrequencyvaluekiiscalculated:wherethesummationcountsthefrequencyofpixelsintheimagewithbrightnessvaluesequaltoorlessthanBVi,andnisthetotalnumberofpixelsintheentirescene(4,096inthisexample).Thehistogramequalizationprocessiterativelycomparesthetransformationfunctionkiwiththeoriginalvaluesofli,todeterminewhichareclosestinvalue.Theclosestmatchisreassignedtotheappropriatebrightnessvalue.
Forexample,weseethatk0
=
0.19isclosesttoL1=0.14.Therefore,allpixelsinBV0(790ofthem)willbeassignedtoBV1.Similarly,the1023pixelsinBV1willbeassignedtoBV3,the850pixelsinBV2willbeassignedtoBV5,the656pixelsinBV3willbeassignedtoBV6,the329pixelsinBV4willalsobeassignedtoBV6,andall448brightnessvaluesinBV5–7willbeassignedtoBV7.Thenewimagewillnothaveanypixelswithbrightnessvaluesof0,2,or4.Thisisevidentwhenevaluatingthenewhistogram.Whenanalystsseesuchgapsinimagehistograms,itisusuallyagoodindicationthathistogramequalizationorsomeotheroperationhasbeenapplied.OriginalHistogramequalizationSpecificpercentagelinearcontraststretchdesignedtohighlightthethermalplumeSpatialFilteringtoEnhanceLow-andHigh-FrequencydetailandEdgesSpatialfrequency,thenumberofchangesinbrightnessvalueperunitdistanceforanyparticularpartofanimage.空間增強(qiáng)SpatialfrequencyinRSimagerymaybeenhancedorsubduedusingspatialconvolutionfiltering
(卷積)
basedprimarilyontheuseofconvolutionmasks.VariousConvolutionMaskKernelsThesizeoftheneighborhoodconvolutionmaskorkernel(n)isusually3x3,5x5,7x7,or9x9.
SpatialConvolutionFiltering:MedianFilterAmedianfilterhascertainadvantageswhencomparedwithweightedconvolutionfilters,including:1)itdoesnotshiftboundaries,and2)theminimaldegradationtoedgesallowsthemedianfiltertobeappliedrepeatedlywhichallowsfinedetailtobeerasedandlargeregionstotakeonthesamebrightnessvalue(oftencalledposterization).SpatialFrequencyFilteringSpatialConvolutionFiltering:MinimumorMaximumFiltersOperatingononepixelatatime,thesefiltersexaminethebrightnessvaluesofadjacentpixelsinauser-specifiedradius(e.g.,3x3pixels)andreplacethebrightnessvalueofthecurrentpixelwiththeminimumormaximumbrightnessvalueencountered,respectively.SpatialFrequencyFilteringWewillconstrainourdiscussionto3x3convolutionmaskswithninecoefficients,ci,definedatthefollowinglocations: c1c2c3Masktemplate= c4c5c6 c7c8c9SpatialConvolutionFiltering111111111Alinear
spatialfilterisafilterforwhichthebrightnessvalue(BVi,j,out)atlocationi,jintheoutputimageisafunctionofsomeweightedaverage(linearcombination)ofbrightnessvalueslocatedinaparticularspatialpatternaroundthei,jlocationintheinputimage.Thecoefficients,c1,inthemaskaremultipliedbythefollowingindividualbrightnessvalues(BVi)intheinputimage: c1xBV1c2xBV2c3xBV3Masktemplate= c4xBV4c5xBV5c6xBV6 c7xBV7c8xBV8c9xBV9TheprimaryinputpixelunderinvestigationatanyonetimeisBV5=BVi,jSpatialConvolutionFiltering111111111SpatialConvolutionFiltering:LowFrequencyFilterSpatialFrequencyFilteringSpatialConvolutionFiltering:HighFrequencyFilterHigh-passfilteringisappliedtoimagerytoremovetheslowlyvaryingcomponentsandenhancethehigh-frequencylocalvariations.Onehigh-frequencyfilter(HFF5,out)iscomputedbysubtractingtheoutputofthelow-frequencyfilter(LFF5,out)fromtwicethevalueoftheoriginalcentralpixelvalue,BV5:SpatialFrequencyFilteringSpatialConvolutionFiltering:Unequal-weightedsmoothingFilter0.250.500.250.5010.500.250.500.25111121111SpatialConvolutionFiltering:EdgeEnhancementFormanyRSEarthscienceapplications,themostvaluableinformationthatmaybederivedfromanimageiscontainedintheedgessurroundingvariousobjectsofinterest.Edgeenhancementdelineatestheseedgesandmakestheshapesanddetailscomprisingtheimagemoreconspicuousandperhapseasiertoanalyze.Edgesmaybeenhancedusingeitherlinearornonlinearedgeenhancementtechniques.SpatialConvolutionFiltering:DirectionalFirst-DifferenceLinearEdgeEnhancementTheresultofthesubtractioncanbeeithernegativeorpossible,thereforeaconstant,K(usually127)isaddedtomakeallvaluespositiveandcenteredbetween0and255SpatialConvolution
Filtering:
High-passFiltersthatAccentuateorSharpenEdges-1-1-1-19-1-1-1-11-21-25-21-21SpatialConvolutionFiltering:
LinearEdgeEnhancement-Embossing00010-1000EmbossEast001000-100EmbossNWSpatialFrequencyFiltering0-10-14-10-10-1-1-1-18-1-1-1-11-21-24-21-211111-71111SpatialConvolutionFiltering:EdgeEnhancementTheLaplacianisasecondderivativeandisinvarianttorotation,meaningthatitisinsensitivetothedirectioninwhichthediscontinuities(point,line,andedges)run.SpatialFrequencyFilteringPrincipalComponentsAnalysis
transformationoftherawremotesensordatausingPCAcanresultinnewprincipalcomponentimagesthatmaybemoreinterpretablethantheoriginaldata.
mayalsobeusedtocompresstheinformationcontentofanumberofbandsofimagery(e.g.,sevenTMbands)intojusttwoorthreetransformedprincipalcomponentimages.Theabilitytoreducethedimensionality(i.e.,thenumberofbands)fromntotwoorthreebandsisanimportanteconomicconsideration,especiallyifthepotentialinformationrecoverablefromthetransformeddataisjustasgoodastheoriginalremotesensordata.光譜增強(qiáng)與變換Thespatialrelationshipbetweenthefirsttwoprincipalcomponents:(a)Scatter-plotofdatapointscollectedfromtworemotelybandslabeledX1andX2withthemeansofthedistributionlabeledμ1andμ2.(b)AnewcoordinatesystemiscreatedbyshiftingtheaxestoanX
system.ThevaluesforthenewdatapointsarefoundbytherelationshipX1=X1–μ1andX2=
溫馨提示
- 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請(qǐng)下載最新的WinRAR軟件解壓。
- 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請(qǐng)聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
- 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會(huì)有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
- 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
- 5. 人人文庫網(wǎng)僅提供信息存儲(chǔ)空間,僅對(duì)用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對(duì)用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對(duì)任何下載內(nèi)容負(fù)責(zé)。
- 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請(qǐng)與我們聯(lián)系,我們立即糾正。
- 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對(duì)自己和他人造成任何形式的傷害或損失。
最新文檔
- 二零二五年度大數(shù)據(jù)處理履行合同安全保密協(xié)議
- 二零二五年度電子商務(wù)運(yùn)營咨詢費(fèi)合同
- 二零二五年度環(huán)保廣告投放與綠色營銷合同匯編
- 二零二五年度供應(yīng)鏈金融終止協(xié)議通知函
- 二零二五年度變壓器制造技術(shù)培訓(xùn)與轉(zhuǎn)讓協(xié)議
- 二零二五年度農(nóng)村安置房租賃保證金及退還合同
- 2025年度校企深度合作人才培養(yǎng)項(xiàng)目協(xié)議書
- 建筑公司勞務(wù)合同(2025年度)勞務(wù)人員工資及福利調(diào)整協(xié)議
- 二零二五年度山東省新建商品房買賣合同預(yù)售與社區(qū)教育服務(wù)協(xié)議
- 二零二五年度高利貸借款合同金融科技賦能發(fā)展
- 蘇科版八年級(jí)物理上冊(cè)同步學(xué)與練第14課光的反射(原卷版+解析)
- 2024年上海市中考語文真題卷及答案解析
- 校園直飲水機(jī)供貨安裝及售后服務(wù)方案
- 廢氣處理系統(tǒng)改造及廢水處理系統(tǒng)改造項(xiàng)目可行性研究報(bào)告
- 小學(xué)全體教師安全工作培訓(xùn)
- 大學(xué)物業(yè)服務(wù)月考核評(píng)價(jià)評(píng)分表
- 現(xiàn)代家政導(dǎo)論-課件 1.1.2認(rèn)識(shí)家政學(xué)起源與發(fā)展
- 期末模擬測試卷(試卷)2024-2025學(xué)年六年級(jí)數(shù)學(xué)上冊(cè)人教版
- 2024屆護(hù)士資格考試必考基礎(chǔ)知識(shí)復(fù)習(xí)題庫及答案(共170題)
- 工業(yè)大數(shù)據(jù)算法賽項(xiàng)實(shí)際操作部分評(píng)分細(xì)則變更說明
- 小學(xué)生防性侵安全教育主題班會(huì)課件
評(píng)論
0/150
提交評(píng)論