打印機(jī)中英文對照外文翻譯文獻(xiàn)

中英文資料翻譯奧普蒂化的光譜紐格伯爾模型打印機(jī)特性

摘要一個比色打印機(jī)需以其注入的一組油墨，預(yù)測由此產(chǎn)生的印刷顏色，通過反射或刺激值指定。該紐格伯爾模式已被廣泛用于預(yù)測的半色調(diào)彩色打印機(jī)色反應(yīng)。本文是介紹和比較優(yōu)化紐格伯爾模型的技術(shù)。這些措施包括尼爾森因素的優(yōu)化，這歸因于光的散射及估計該點區(qū)域并擴(kuò)展到精確位置的功能。一種新的技術(shù)描述了優(yōu)化紐格伯爾初選使用加權(quán)譜回歸。實驗結(jié)果提出了它使用兩個半色調(diào)屏幕：隨機(jī)靜電打印機(jī)或旋轉(zhuǎn)點，和互聯(lián)網(wǎng)上點屏幕。使用尤爾一尼爾森因子，細(xì)微的的框架，并使譜回歸模型的準(zhǔn)確度大大提高。1引言比色打印機(jī)特性要求是建立在信號的輸入到打印機(jī)和由此產(chǎn)生的印刷顏色的色度測量的關(guān)系上的。這種關(guān)系，我們稱之為打印機(jī)特性的功能，往往是得到了印刷和測量補(bǔ)丁的色彩大量運用了一些在已知的樣本的測量插值。另一種方法近似于用打印機(jī)型號的特征函數(shù)。試圖模擬打印機(jī)經(jīng)常會產(chǎn)生有價值的直覺的基本物理過程，它在光，色彩和紙之間產(chǎn)生復(fù)雜的相互作用。模型的實際優(yōu)點是打印機(jī)可以實現(xiàn)定性并相對較少的測量。打印機(jī)的模型精度顯然取決于模型中的假設(shè)是否正確。該紐格伯爾模型已廣泛用于二進(jìn)制模式的彩色打印機(jī)并可采用各種色調(diào)屏幕。原始模型實質(zhì)上是Murray-Daviesequation（用于預(yù)測灰度反射）的色彩方法。打印機(jī)字的顏色由紅綠藍(lán)（RGB）三色組合，預(yù)計作為對固體套印的RGB值加權(quán)平均印刷；青色，品紅和黃色的印刷，由地區(qū)覆蓋范圍相對點的權(quán)重決定。在這里，R，G，B可能被認(rèn)為是從印刷補(bǔ)丁反射光，通過3個已知光譜靈敏度的職能通過寬帶過濾器，超過估計所有在五個可見光波長范圍內(nèi)的方法，從該地區(qū)的覆蓋范圍數(shù)字輸入色調(diào)，包括一些直接測量和計算的結(jié)合。請注意，盡管這種模式提供設(shè)備值（CMY）的特征函數(shù)（三原色）色值，它是從色值到設(shè)備空間，最終是為了尋求色彩再現(xiàn)逆映射。幾位研究人員討論了反相的紐格伯爾方程。這個反演問題是一個非線性問題，可以解決在三個輸入著色劑的情形C，M，Y，通過迭代方法。當(dāng)著色劑數(shù)量超過3，問題就變得病態(tài)數(shù)著色劑組合可以導(dǎo)致生成相同的測量顏色。在這項工作中，只有前進(jìn)定性問題的審查。筆者認(rèn)為，一旦提出一個精確的模型，樣品的高密度，導(dǎo)出等多層面的擬合和插值技術(shù)可以很容易被用來顛倒模式。各種方法已經(jīng)被提出了提高原紐格伯爾方程的準(zhǔn)確性。一個重要的現(xiàn)象，不占原模型，光散射的文件內(nèi)。YuleandNielsen通過模型中的一個額外的參數(shù)方程紐格伯爾這種效果。其他研究人員提出了在之間的透射模式和一紙點擴(kuò)散函數(shù)（PSF）的回旋空間光散射的形式更復(fù)雜的模型。尼爾森參數(shù)不明確的空間變化之間的相互作用光紙帳戶，有人建議由魯克德舍爾和豪澤，一個簡單的關(guān)系可能與這個簡單的模型和滌綸短纖存在的模型。維賈諾表明，而不是在紐格伯爾方程形式的收益率更高的精度寬帶頻譜。這一發(fā)現(xiàn)證實了作者。Engeldrum表明，無論是紙張和圓點的顏色相對點區(qū)域覆蓋范圍內(nèi)的功能，并延長了紐格伯爾模型交代。利用該紐格伯爾方程的幾何解釋，開發(fā)了蜂窩Hueberger盧哥貝爾模式，即除固體套印顏色來計算最終輸出的顏色。蜂窩技術(shù)將被視為在本文件。從德米歇爾的假設(shè)出發(fā)點區(qū)域覆蓋范圍內(nèi)，并用序列二次規(guī)劃方法來估計這些參數(shù)。探討了波長依賴尤爾使用尼爾森的因素。在這項工作中采取的做法是，以符合理想打印機(jī)的紐格伯爾模型的基本形式，并提供優(yōu)化模型的參數(shù)，以便更好地適應(yīng)實際打印機(jī)的特性的數(shù)學(xué)技術(shù)。因此，該模型簡單保留，而其精度提高。優(yōu)化的圣誕-尼爾森因素，技術(shù)點區(qū)域功能，細(xì)胞模型已被提交會議的出版物上發(fā)表。本文介紹這些技術(shù)在一個比較一致的框架。此外，由在最近的一次會議報告的作者提出譜回歸的新方法，更詳細(xì)。實驗結(jié)果提出了兩種類型的半色調(diào)屏幕：隨機(jī)或旋轉(zhuǎn)點，和互聯(lián)網(wǎng)上點屏幕上。2紐格伯爾混合模型。二進(jìn)制打印機(jī)，一個多種顏色的渲染，是通過不同的組合達(dá)到點著色劑的主要地區(qū)覆蓋范圍。在紐格伯爾方程'預(yù)測從任意作為一種已知的基礎(chǔ)上為2n的色彩，光譜反射加權(quán)平均印刷補(bǔ)丁的平均光譜反射率，其中N是系統(tǒng)著色劑數(shù)量。在這項工作中的N=3CMY打印機(jī)案件將用來描述各種模型。方程很容易推廣到N著色劑，而事實上，青色，洋紅色，黃色，黑色（CMYK）打印機(jī)將在實驗結(jié)果中。對于CMY打印機(jī)，有8個的基礎(chǔ)上的顏色，這是白皮書（寬）和1，2和3純色覆蓋（即0％和C，男，y的100％的組合），這些基礎(chǔ)色將稱為紐格伯爾初選。在原來的紐格伯爾方程，預(yù)測的顏色是由三個寬帶代表短期，中期反射和電磁波譜的長波長部分指定。在這項工作中，窄帶反射光譜住宅（十）使用，而不是他們的寬帶同行，因為前者一般收益率更高的準(zhǔn)確性。3結(jié)論

本文幾個打印機(jī)的紐格伯爾方程的模型已被描述和比較。一種新的技術(shù)已提出了優(yōu)化光譜紐格伯爾初選使用最小二乘回歸。提供各種型號的復(fù)雜性和精確度之間不同的權(quán)衡。表1總結(jié)了在測量數(shù)量而言，這些權(quán)衡需要制定出一個給定的模式和結(jié)果的準(zhǔn)確性。一般來說，模型的準(zhǔn)確度是成正比的測量數(shù)目，但人們可以方便地識別報酬遞減的情況。產(chǎn)生的巨大收益來自使用尤爾一尼爾森改正，并可以在細(xì)胞框架或全球譜回歸模型。測量的數(shù)目增加的細(xì)胞模型與細(xì)胞數(shù)量成倍的要求，同時在準(zhǔn)確增益減小。在34=81初選的情況常常是一個可以接受的折衷。在當(dāng)?shù)?，在計算成本的大幅增加加?quán)回歸結(jié)果（未顯示在表1），在沒有獲得準(zhǔn)確對全球回歸。在執(zhí)行回歸的數(shù)學(xué)解釋是，它提供了一個自由的模式（即初選額外的程度），因此適合優(yōu)越的經(jīng)驗數(shù)據(jù)。從初選的光譜，如無花果，陰謀的研究。10和11，已經(jīng)從沒有測量，觀察到的趨勢回歸職能。因此，很難物理意義重視對這些陰謀，除了指出，在選定的樣本良好的線性回歸可能會導(dǎo)致的光譜測量噪聲平均，從而得到更強(qiáng)大的預(yù)測。這項工作提出了一些進(jìn)一步的擴(kuò)展。譜回歸應(yīng)用于蜂窩架構(gòu)可能會導(dǎo)致沒有額外的提高測量精度。阿譜初選的聯(lián)合優(yōu)化，點區(qū)域功能和尤爾一尼爾森因素可能導(dǎo)致對其中n因素是單獨優(yōu)化目前的做法更好的結(jié)果。這是身體有理由認(rèn)為在光散射波長基板依賴，因此一個譜回歸模型，支持n是X的函數(shù)可能會導(dǎo)致更準(zhǔn)確的模型的擴(kuò)展。然而，重要的是銘記之間的物理模型的復(fù)雜性和準(zhǔn)確性實現(xiàn)權(quán)衡始終牢記實際打印系統(tǒng)。最后，這將是值得研究的印刷技術(shù)比其他如靜電復(fù)印，噴墨機(jī)，膠印機(jī)等更具有效性。OptiizationofthespectralNeugebauermodelforprintercharacterizationAbstractAcolorimetricprintermodeltakesasitsinputasetofinkvaluesandpredictstheresultingprintedcolor,asspecifiedbyre?flectanceortristimulusvalues.TheNeugebauermodelhasbeenwidelyusedtopredictthecolorimetricresponseofhalftonecolorprinters.Inthispaper,techniquesforoptimizingtheNeugebauermodelarepresentedandcompared.TheseincludeoptimizationoftheYule-Nielsenfactorthataccountsforlightscatteringinthepa?per,estimationofthedotareafunctions,andextensiontoacellularmodel.AnewtechniqueisdescribedforoptimizingtheNeugebauerprimariesusingweightedspectralregression.Experimentalresultsarepresentedforxerographicprintersusingtwohalftonescreens:therandomorrotateddot,andthedot-on-dotscreen.UseoftheYule-Nielsenfactor,thecellularframework,andspectralregressionconsiderablyincreasemodelaccuracy.?1999SPIEand1S&T.1IntroductionColorprintercharacterizationrequiresthatarelationshipbeestablishedbetweentheinputsignalstotheprinterandthecolorimetricmeasurementsoftheresultingprintedcolors.Thisrelationship,whichwecalltheprintercharacterizationfunction,isoftenobtainedbyprintingandmeasuringalargenumberofcolorpatchesandapplyingsomeinterpo?lationamongthemeasurementsattheknownsamples.Analternativeapproachistoapproximatethecharacterizationfunctionwithaprintermodel.Anattempttomodeltheprinteroftenyieldsvaluableintuitionabouttheunderlyingphysicalprocess,whichconstitutescomplexinteractionsbetweenthelight,colorant,andpaper.Apracticaladvan?tageofmodelingisthattheprintercharacterizationcanbeachievedwitharelativelysmallnumberofmeasurements.Theaccuracyoftheprintermodelclearlydependsonthevalidityoftheassumptionsinthemodel.TheNeugebauermodelhasbeenwidelyusedtomodelbinarycolorprintersemployingvarioushalftonescreens.TheoriginalmodelisessentiallyanextensionoftheMurray-Daviesequation2(usedforpredictinggrayscalere?flectance)tothecolorcase.Thecolorofaprint,inred?green-blue(RGB)coordinates,ispredictedasaweightedaverageoftheRGBvaluesofthesolidoverprintsofthethreeprintingprimaries,cyan,magenta,andyellow(C,M,Y),wheretheweightsaredeterminedbytherelativedotareacoveragesc,m,yconstitutingtheprint.Here,R,G,Bmaybethoughtofasthereflectedlightfromtheprintedpatch,passedthroughthreebroadbandfilterswithknownspectralsensitivityfunctions,andsummedoverallwave?lengthsinthevisiblerangeV.Methodsforestimatingdotareacoveragesfromthedigitalinputvaluestothehalftoneincludesomecombinationofdirectmeasurementandcal?culation.Notethatwhilethemodelprovidesthecharacterizationfunctionfromdevicevalues(CMY)tocolorimetricvalues(RGB),itistheinversemappingfromcolorimetrictode?vicespacethatisultimatelysoughtforcolorreproduction.Severalresearchershaveaddressedtheproblemofinvert?ingtheNeugebauerequations.3-6Theinversionisanonlin?earproblemthatcanbesolved,inthecaseofthreeinputcolorantsC,M,Y,byiterativeapproaches.Whenthenum?berofcolorantsexceedsthree,theproblembecomesill-posed,asseveralcolorantcombinationscanresultinthesamemeasuredcolor.Inthiswork,onlytheforwardchar?acterizationproblemisconsidered.Itistheauthor'sopin?ionthatonceasufficientlydensesamplingofanaccurateforwardmodelisderived,othermultidimensionalfittingandinterpolationtechniques7?8caneasilybeusedtoinvertthemodel.VariousmethodshavebeenproposedforimprovingtheaccuracyoftheoriginalNeugebauerequations.Animpor?tantphenomenonnotaccountedforintheoriginalmodelisthatoflightscatteringwithinthepaper.YuleandNielsen9modeledthiseffectviaanadditionalparameterintheNeu?gebauerequations.Otherresearchers10-12haveproposedmoresophisticatedmodelsforlightscatteringintheformofaspatialconvolutionbetweenthetransmittancepatternandapaperpointspreadfunction(PSF).WhiletheYule-Nielsenparameterdoesnotexplicitlyaccountforspatiallyvaryinginteractionsbetweenlightandpaper,ithasbeensuggestedbyRuckdeschelandHauser13thatasimplerela?tionshipmayexistbetweenthissimplemodelandthePSFbasedmodels.Viggiano14,15demonstratedthatworkingwiththespectralratherthanthebroadbandformoftheNeugebauerequationsyieldshigheraccuracy.Thisfindinghasbeenconfirmedbytheauthor.16Engeldrumeta/.17.I'showedthatthecolorsofboththepaperandthedotsarefunctionsoftherelativedotareacoverages,andextendedtheNeugebauermodeltoaccountforthis.UtilizingthegeometricinterpretationoftheNeugebauerequations,Hueberger19developedthecellularNeugebauermodel,wherecolorsinadditiontothesolidoverprintsareusedtocalculatethefinaloutputcolor.Thecellulartechniquewillbeconsideredinthispaper.Leeetal.20departedfromtheDemichelassumptionsondotareacoverages,andusedasequentialquadraticprogrammingmethodtoestimatetheseparameters.HuaandHuang21andIinoandBems22."ex?ploredtheuseofawavelength-dependentYule–Nielsenfactor.TheapproachtakeninthisworkistoconformtothebasicformoftheNeugebauermodelforanidealprinter,andtooffermathematicaltechniquesforoptimizingtheparametersofthemodeltobetterfitthecharacteristicsofrealprinters.Thusthesimplicityofthemodelisretained,whileitsaccuracyisimproved.TechniquesforoptimizingtheYule–Nielsenfactor,dotareafunctions,andcellularmodelhavebeenpublishedinpreviousconferencepublica?tionsbytheauthor.16'24Inthispaper,thesetechniquesarepresentedandcomparedinacoherentframework.Inaddi?tion,anewmethodofspectralregressionproposedbytheauthorinarecentconferencereport25ispresentedingreaterdetail.Experimentalresultsarepresentedfortwotypesofhalftonescreens:therandomorrotateddot,andthedot-on?dotscreen.Thepaperisorganizedasfollows.InSec.2,theNeuge?bauermixingmodelisbrieflyintroduced,andisappliedtothetwotypesofhalftoneconfigurations.Thelightscatter?ingproblemandYule–NielsencorrectionaredescribedinSec.3.MethodsforcalculatingthedotareafunctionsaredescribedinSec.4,andextensionstothecellularframe?workarepresentedinSec.5.Aspectralregressiontech?niqueisdescribedinSec.6,wheretheNeugebauerprima?riesarethemselvestreatedasparameterstobeoptimized.ExperimentalresultsarepresentedinSec.7,andconclud?ingremarksarecollectedinSec.8.2NeugebauerMixingModelorabinaryprinter,therenderingofamultitudeofcolorsisachievedthroughcombinationsofvaryingdotareacov?eragesoftheprimarycolorants.TheNeugebauerequations'predicttheaveragespectralreflectancefromanarbitraryprintedpatchasaweightedaverageofthespectralreflectancesof2Nknownbasiscolors,whereNisthenum?berofsystemcolorants.Inthiswork,thecaseofN=3forCMYprinterswillbeusedtodescribethevariousmodels.TheequationsareeasilygeneralizedtoNcolorants;andindeed,cyan–magenta–yellow–black(CMYK)printerswillbeincludedintheexperimentalresults.2.3EffectofScreenDesignOneofthefundamentalfactorsthataffectsmodelaccuracyisthefrequencyofthehalftonescreen.".28Theidealmodelassumesthatthedotsarerectangularincrosssection,andhenceresultinabinaryabsorptionprofileasafunctionofspatiallocation.Inreality,dotshavesofttransitionsfromregionswithfullinktoregionswithnoink.Ifthehalftonescreenfrequencyisrelativelylow,oraclustereddotisused,therelativeareaofthepapercoveredbythetransitionpectedtoberelativelyaccurate.Ontheotherhand,ifthescreenfrequencyishigh,oradisperseddotisused(asisthecasewithaBayerscreenorerrordiffusion),thenalargefractionofthepaperiscoveredbytransitoryregions,andthemodelbreaksdown.ThecorrectionsdiscussedinSec.3partiallyaccountforsofttransitions;nevertheless,thereliabilityofthemodelhasbeenseentobegreatestwithclus?tereddotscreensintherangeof300-400halftonedotsperinch.8ConclusionsInthispaper,severalprintermodelsbasedontheNeuge?bauerequationshavebeendescribedandcompared.AnewtechniquehasbeenproposedforoptimizingthespectralNeugebauerprimariesusingleastsquaresregression.Thevariousmodelsofferdifferenttradeoffsbetweencomplex?ityandaccuracy.Table1summarizesthesetradeoffsintermsofthenumberofmeasurementsrequiredtoderiveagivenmodelandtheresultingaccuracy.Generally,modelaccuracyisproportionaltothenumberofmeasurements;howeveronecaneasilyidentifycasesofdiminishingre?turns.ThesignificantgainscomefromusingtheYule-Nielsencorrection,andeitherthecellularframeworkortheglobalspectralregressionmodel.Thenumberofmeasure?mentsrequiredinthecellularmodelincreasesexponen?tiallywiththenumberofcells,whilethegaininaccuracydiminishes.Thecaseof34=81primariesisoftenanac?ceptabletradeoff.Thelocallyweightedregressionresultsinasignificantincreaseincomputationalcost(notshowninTable1),withlittlegaininaccuracyovertheglobalregres?sion.Themathematicalinterpretationofperformingregres?sionisthatitaffordsanextradegreeoffreedominthemodel