資源描述：
《Robust Real-Time Face Detection》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、InternationalJournalofComputerVision57(2),137–154,2004c2004KluwerAcademicPublishers.ManufacturedinTheNetherlands.RobustReal-TimeFaceDetectionPAULVIOLAMicrosoftResearch,OneMicrosoftWay,Redmond,WA98052,USAviola@microsoft.comMICHAELJ.JONESMitsubishiElectricResearch
2、Laboratory,201Broadway,Cambridge,MA02139,USAmjones@merl.comReceivedSeptember10,2001;RevisedJuly10,2003;AcceptedJuly11,2003Abstract.Thispaperdescribesafacedetectionframeworkthatiscapableofprocessingimagesextremelyrapidlywhileachievinghighdetectionrates.Therearethr
3、eekeycontributions.The?rstistheintroductionofanewimagerepresentationcalledthe“IntegralImage”whichallowsthefeaturesusedbyourdetectortobecomputedveryquickly.Thesecondisasimpleandef?cientclassi?erwhichisbuiltusingtheAdaBoostlearningalgo-rithm(FreundandSchapire,1995)
4、toselectasmallnumberofcriticalvisualfeaturesfromaverylargesetofpotentialfeatures.Thethirdcontributionisamethodforcombiningclassi?ersina“cascade”whichallowsback-groundregionsoftheimagetobequicklydiscardedwhilespendingmorecomputationonpromisingface-likeregions.Aset
5、ofexperimentsinthedomainoffacedetectionispresented.Thesystemyieldsfacedetectionperfor-mancecomparabletothebestprevioussystems(SungandPoggio,1998;Rowleyetal.,1998;SchneidermanandKanade,2000;Rothetal.,2000).Implementedonaconventionaldesktop,facedetectionproceedsat1
6、5framespersecond.Keywords:facedetection,boosting,humansensing1.Introductionencesinvideosequences,orpixelcolorincolorim-ages,havebeenusedtoachievehighframerates.OurThispaperbringstogethernewalgorithmsandinsightssystemachieveshighframeratesworkingonlywithtoconstruc
7、taframeworkforrobustandextremelyrapidtheinformationpresentinasinglegreyscaleimage.visualdetection.TowardthisendwehaveconstructedThesealternativesourcesofinformationcanalsobein-afrontalfacedetectionsystemwhichachievesdetec-tegratedwithoursystemtoachieveevenhigherf
8、rametionandfalsepositiverateswhichareequivalenttorates.thebestpublishedresults(SungandPoggio,1998;Therearethreemaincontributionsofourfacedetec-Rowleyetal.,1998
