颜色识别中英文对照外文翻译文献.docx

颜色识别中英文对照外文翻译文献.docx

《颜色识别中英文对照外文翻译文献.docx》由会员分享，可在线阅读，更多相关《颜色识别中英文对照外文翻译文献.docx（14页珍藏版）》请在冰点文库上搜索。

颜色识别中英文对照外文翻译文献

中英文资料对照翻译

英文原文

AnApproachofColorFeatureEvaluationinColor

Abstract—Thispaperanalyzesthecharacteristicsoffivecommonlyusedcolorspacesandexplorestheirinfluencesoncolorrecognitionrespectively.DivisibilityevaluationbasedondistancecriterionisutilizedtoevaluatethedifferentcolorfeaturesineachcolorspaceandexperimentalresultsshowthatHSIcolorspacehasthebestdivisibilityperformance.Keywords-colorspace;colorrecognition;featureevalutation;divisibilitycritiron

I.INTRODUCTION

Coloristhemostintuitivevisionfeaturetodescribecolorfulimages.Ithasbeenwidelyusedinpatternrecognitionforthereasonthatcolorfeatureisalmostfreefromtheeffectsofscale,rotationandtranslationfortheinputimages[1].Colorsincolorfulimagescanbedefinedbydifferentcolorspacemodels,suchasRGBspace,CMYspace,I1I2I3space,YUVspaceandHSIspace.Amongtheabovecolorspaces,RGBisthebasicandthemostcommononeandcanreadilybemappedintoothercolorspaces.However,RGBspaceisnon-uniformfor

colorperceptionandistooeasilyinfluencedbylight.ThethreecolorcomponentsofRGBspacearecorrelatedwitheachother[2].CMYspacerepresentscolorsbythecomplementarycomponentofRGBcomponents.YUVspace,frequentlyusedincolorTVsystems,usesthreechannelsasY,UandVtodefinethepixel.Yarethebrightnessinformation,UandVarethecolordifferencewhichdenotestheoverallcolordifferenceinsteadofthedifferencebetweenthethreecomponentsofRGB.HSIspaceisauniformonewhichconsiststothehumanperceptionofcolors.Itsthreecomponentsaremutuallyindependentandcanperceivecolorchangeofeachcomponentrespectively.Butnon-lineartransforminHSIspacemayleadsubstantialcomputationaswellassingularityofthecolorspacewhenthesaturationislow.WhileinYCbCrcolorspace,thechrominancecomponentandtheluminancecomponentareinterdependent.Besidesthat,theconversionfromYCbCrspacetoRGBspaceislinearandsimple,soYCbCrspaceiscommonlyusedinthefieldofvideoencodingcompression.YUVspace,YCbCrspaceandHSIspaceallrepresentspectrumintwodimensionandusethethirddimensiontorepresenttheintensity

ofcolor,whichenablesthemmoresuitableforoccasionswherelightintensitychanges,thanRGBspace.Colorrecognitiontechniquehasbeenappliedtomanyfieldsandhasgoneaheadrapidly.Forinstance,colorrecognitioninproductsurface,licenseplatesidentification,facerecognitionandskinrecognition[3-6].Colorrecognitioneffectsdifferwiththechangeofcolorspace.ThispaperinvestigatesoncolorfeaturedivisibilityinthecommonlyusedcolorspacesasRGBspace,CMYspace,YUVspace,YCbCrspace,I1I2I3spaceandHSIspace.AnalysisindicatesthatHSIhasthebestdivisibilityperformanceinallthe

abovecolorspacesbasedonthedistancecriterion.Itprovidesatheorybasisforcolorrecognition.

II.COLORSPACEANDITSTRANSFORMATION

Itisessentialtobuildupandselectasuitablecolorspaceforobtainingakindofvalidcolorfeaturestocharacterizecolorfulimages.Differentcolorspacesareutilizedfordifferentresearchpurposes.Colorspacemeanstodefinecolorbyan

arrayinthree-dimensionspace.Intheprocessingofcolorfulimages,colorspaceisalsonamedascolormodelorcolorcoordinates.Onecolorspacecanbeconvertedtoanotherbycertaintransforms.Belowistheintroductionofsomecolorspacesandtheirconversions[7].

A.RGBColorSpace

Red（R）,green（G）,blue（B）arethreeprimarycolorsofspectrum.Allcolorscanbegeneratedbythesumofthethreeprimarycolors.Indigitalimages,valuesofR,GandBrangefrom0to255.Acubeinthree-dimensioncoordinatespacecanbeusedtodescribetheRGBcolorspace,wherered,greenandbluearethethreeaxes,showninFig.

1.ThemaindrawbackofRGBcolorspaceasfollows:

•Itisnotintuitive.ItisdifficulttoseefromtheRGBvaluesthecognitiveattributesthatthecolorrepresentsitself.

•Itisnon-uniform.TheperceptiondifferencebetweentwocolorsinRGBspaceisdifferentfromthedistancebetweenthetwocolors.

•Itisdependentonhardwaredevices.

Inaword,RGBspaceisdevice-relatedandanincompleteintuitivecolordescription.Toovercometheseproblems,othercolorspaces,whicharemoreinlinewithcharacteristicsofcolorvision,areadopted.RGBspacecanbemappedtoothercolorspacesreadily.

B.CMY（CMYK）ColorSpace

CMYspaceisaspatialstructureofarectangularCartesian.Itsthreeprimarycomponentsarecyan（C）,magenta（M）andyellow（Y）.Colorsareobtainedbysubtractivecolors.CMYspaceiswidelyusedinnon-emissiondisplayasinkjetprinters.Equalamountofthethreecomponentscangeneratetheblackcolor.Buttheaforementionedblackcolorisnotpure.Generallyspeaking,togeneratetrueblackcolor,thefourthcomponent,i.e.black,isaddedin.ThisistheCMYKcolorspace.CMYspaceisnotveryintuitiveandnon-linear.ItsthreecomponentsarethecomplementarycolorsofR,GandB.Thetransformationsareasfollows:

ThetransformationsfromRGBspacetoCMYspaceareasfollows:

C.YUVandYCrCbColorSpace

YUVspaceandYCbCrspacebothgeneratealuminancecomponentandtwo

chrominancecomponents.InYUVspace,Yistheluminancecomponent,Uand

Vcolordifference.Ycomponentisindependentoftheothertwo.Moreover,theYUVspacecanreducethestoragecapacityrequiredbydigitalcolorfulimagesbythecharacteristicofhumanvision.InYCbCrspace,Yistheluminancecomponent,CbisthebluecolorcomponentandCrtheredcolorcomponent.ItsadvantagesareobviousthatcolorcomponentsareseparatedfromluminancecomponentsandlineartransformationcanbeperformedfromRGBspace.TransformationfromRGBspacetoYUVspacecanbeexplainedapproximatedbythefollowingequations:

D.HSIColorSpace

HSIspaceisestablishedfromthehumanpsychologicalperceptionpointofview.H（hue）isacolorinacolorcorrespondingtothemainwavelengthinchromatography.S（saturation）isequivalenttothepurityofcolor.I（intensity）isthebrightnessofcolorandtheuniformamountoffeeling.HSV（hue,saturation,value）andHSB（hue,saturation,brightness）areothercolorspacessimilartoHSIcolorspace,andareallbelongtopolarcoordinatespacestructure.Theircommonmeritisthattheycandescribethecolorintuitively.MostofthemcanbeconvertedfromRGBspacelinearly.HSIcolorspacehastwoimportantpoints.OneisthatIcomponentisseparatedfromHcomponent,i.e.Icomponentisindependentofimagecolorinformation.TheotheristhatHcomponentandScomponentarecloselylinkedtothewayhumanfeelingcolor,wherethecolordescriptionabilityofHcomponentisthemostclosettohumanvision.AndthendistinguishabilityofHcomponentisthestrongest[8].TransformationfromRGBspacetoHSIspacecanbeexplainedbythefollowingequations:

HSIcolorspaceprovidesasuitablespacewiththreecomponentsthatisbetterusedtodescriptcolorinlinewithhumanhobbits.However,thedefectofnon-linearincolordifferencestillexists,especiallythecolorandangleintheHcomponent[9].

E.I1I2I3ColorSpace

LineartransformationfromRGBspacetoI1I2I3spacecanbeexplainedbythefollowingequationstogetthreeorthogonalcolorfeatures:

Fromformula（7）,itcanbeseenthatvaluesofI1,I2andI3componentcanbepositiveandnegative.Thenon-correlationpropertyofI1I2I3spaceisthebestinimagerecognition.

III.FEATUREEVALUATIONOFCOLORSPACE

Bycolorspaces,theabstract,subjectivevisualperceptioncanbetranslatedintoaconcretespecificposition,vectorinthree-dimensionalspace,whichmakesitpossibletovisualizecolorfeaturesofcolorfulimagesanddevices.Colorspaceisanimportanttoolofcolorrecognition.Variousmixingsystemhasitscorrespondingcolorspace,anddifferentcolorspaceshavedifferentpropertieswiththeirrespectiveadvantagesanddisadvantages.Validityofcolorspaceisthekeytocolorimageprocessing.Divisibilitycriterioncanbeusedtotestdifferentcolorspacefortheirperformanceoncolorclassification.Thedistancecriterioniswidelyutilizedduetoitsconciseandclearconcept.Itsprincipleisthatthesmallerthedistancewithinaclasswhilethegreaterthedistancebetweenclasses,thebetterthedivisibilityithas.Belowisthepresentedalgorithmoffeatureevaluationbasedondistancecriterion[10].

•Calculatethemeanvectorandcovarianceoftheithclasssamples,NisthenumberoftotalsamplesandNithenumberoftheithclasssamples.

IV.EXPERIMENTALRESULTSANDANALYSIS

Whenidentifiedbyhumaneyes,colorsaredividedintoelevencategoriesasred,green,blue,yellow,purple,orange,pink,brown,gray,whiteandblack,showninFig.2.

TheevaluationalgorithmisperformedrespectivelyonRGBspace,CMYspace,YUVspace,I1I2I3spaceandHSIspace.FeatureparameterandassessmentindicatorsareshowninTableⅠ.

SeenfromTableⅠ,theHSIspacehasthebestperformancecomparedtootherfouranalyzedcolorspaces.

V.CONCLUSION

Itisnecessarytoselectaneffectivecolorspaceforcolorfulimageprocessing.Thispaperanalyzesandcomparestheperformanceoffivecommoncolorspacesbasedondivisibilitycriterion.ExperimentalresultsshowthatHSIcolorspacehasthebestdivisibilityperformancecomparedwiththeotherfourspaces.Itprovidesabasisforcolorspaceselectionincolorrecognition.

ACKNOWLEDGMENT

ThisworkwassupportedbyagrantfromNaturalScienceFoundationofLiaoningProvince（Grantnumber:

20102153）.

REFERENCES

[1]WANGHui,LVYan,ZHANGKa,“Research