人脸识别系统中的特征提取

:10001506(2001)02001804李华胜,杨桦,袁保宗(,100044):研究了人脸识别系统中正面人脸的特征提取.通过区域增长从人脸图像中分割出人脸,再利用边缘检测Hough变换模板匹配和方差投影技术可以快速有效地提取出人脸面部器官眼睛鼻子和嘴巴特征.实验结果表明本文所采用的方法具有较高的准确率和光照鲁棒性.:人脸识别;Hough变换;模板匹配;方差投影:TP391.4:AFeatureExtractioninHumanFaceRecognitionSystemLIHuasheng,YANGHua,YUANBaozong(CollegeofComputerandInformationTechnology,NorthernJiaotongUniversity,Beijing100044,China)Abstract:Thispaperdealswithfrontalfacialfeatureextractioninfacerecognitionsystem.Firsthumanfaceissegmentedfromtheinputimagebyusingregiongrowth.Thenbyusingedgedetection,Houghtransformation,templatematchingandvarianceprojection,facialfeaturessuchaseyes,noseandmoutharelocatedquicklyandeffectively.Experimentresultsshowthatthemethodwepresenthasahigheraccuracyandrobustperformancetoillumination.Keywords:humanfacerecognition;Houghtransformation;templatematching;varianceprojection,[1~3].,,3:.,.4:.3:[4];[5];![6].,:,,.[7].,..∀#,,.,Hough,,.:20001220;:20010320:(69789301):(1977∃),,,.email:[e-mail]25220014JOURNALOFNORTHERNJIAOTONGUNIVERSITYVol.25No.2Apr.20011,,..,..g(x,y)=a+ln(f(x,y)+1)blnc(1)(a)(b)1,f(x,y),g(x,y),a,b,c.,,.3%3,.,(1).221.NRi,Ris(i,j),Ci1,G.s(i,j)8G=|s(i,j)-s(i-1,j)|(2)G,Ci1,,2N.,,,,.,,.22(1),,,.2.2I(x,y)(x,y),[y1,y2]V(x)[x1,x2]H(y)V(x)=1y2-y1&y2y1I(x,y)H(y)=1x2-x1&x2x1I(x,y)(3),.3,3.V(x),.[8]2v(x)=1y2-y1&y2yi=y1[I(x,yi)-V(x)]2(4)192:2h(y)=1x2-x1&x2xi=x1[I(xi,y)-H(x)]2(5)32v(x).3,,.,4.,,P1,P2,P3,P44.,P1,P2,P3,P4,.34(2)Prewitt,Hough.256%256,,[3,5].,,.,(5)..,e(i,j);,.,1,e(i,j)=1,&(iin+1,jin+1)(i,j)=(iin-1,jin-1)f(i,j)15%&(iout+1,jout+1)(i,j)=(iout-1,jout-1)f(i,j)0,otherwise(6)5(iin,jin)=(i-2cos,j-2sin)(7)(iout,jout)=(i+2cos,j+2sin)(8),f(i,j),(iin,jin)(iout,jout),(i,j).R,Rc(i,j)R=&(i,j)∋c2e(i,j)&(i,j)∋c5e(i,j)e(i,j)=1(9),c2c5,25,5,(6).,,.(3),,,.,,.,.,.,,.,,..,,.,2025,(6).(4),,,,.,,.,,.,,,.,,,,.,,,..,,.,(6).63PII400PC,CCD256%25615,12.:,,155,25,86.1%;172,8,95.5%.,.8,.:[1]ZhangJun,YanYong,MartinLades.FaceRecognition:Eigenface,ElasticMatching,andNeuralNets[J].ProceedingsoftheIEEE,1997,85(9):1422-1435.[2]SamalA,IyengarPA.AutomaticRecognitionandAnalysisforHumanfacesandFacialExpressions:ASurvey[J].PatternRecognition,1992,25(1):65-77.[3]AraV.Nefian.StatisticalApproachestoFaceRecognition[R].GeorgiaInstituteofTechnagySchodofElectricalEngineering,1996.12.[4]HongZiquan.AlgebraicFeatureExtractionofImageforRecognition[J].PatternRecognition,1991,24(3):211-219.[5]LeeSY,HamYK,ParkRH.RecognitionofHumanFrontFacesUsingKnowledgebasedFeatureExtractionandNeuroFuzzyAlgorigthm[J].PatternRecognition,1996,29(11):1863-1876.[6]ChowG,LiX.TowardsASystemforAutomaticFacialFeatureDetection[J].PatternRecognition,1993,26(12):1739-1775.[7]YuilleAL.DetectionTemplatesforFaceRecognition[J].CognitiveNeuroscience,1991,3(1):59-70.[8]FengGC,YuenPC.VarianceProjectionFunctionandItsApplicationtoEyeDetectionforHumanFaceRecognition[J].PatternRecognition,1998,19:899-906.212: