基于Gabor小波的人脸特征提取算法研究及仿真

西南科技大学本科生毕业论文1Southwestuniversityofscienceandtechnology本科毕业设计（论文）题目名称：基于Gabor小波的人脸特征提取算法研究及仿真学院名称专业名称学生姓名学号指导教师二〇一二年六月西南科技大学本科生毕业论文2基于Gabor小波的人脸特征提取算法研究及仿真摘要：人脸识别技术是模式识别与人工智能的研究热点之一。在生物特征识别中，人脸识别占有极为重要的地位。它在访问控制、司法应用、电子商务和视频监控等领域都有广泛的应用。人脸特征提取是人脸识别过程的核心，特征提取的有效性直接影响到分类的速度和识别的性能。本论文的主要研究内容包括以下方面：(1)基于Gabor变换的人脸特征提取算法。通过对人脸傅里叶变换和Gabor变换的实验比较，证明了Gabor变换在提取人脸特征方面具有很大的优越性。接着，介绍了Gabor小波变换的快速算法。(2)针对Gabor小波的缺点，引进LBP算子，该算子的使用能有效的克服Gabor小波变换的缺点。另外还详细介绍了如何提取“特征脸”的原理和实现过程，对Gabor滤波器参数的选择问题也作了详细说明。(3)针对Gabor和LBP提取的人脸特征向量维数过高问题，本文分别采用了LPP和PCA降维算法来进行降维。(4)对最后提取的人脸特征通过计算特征向量的距离来量化查询图像和图像库中每幅图像间的相似程度。该部分主要是将提取出来的人脸特征向量进行验证，并对两种不同降维方法做比较。本文最后是在orl和yale两个人脸图库做仿真实验。实验表明，采用用LPP降维得到的相似度要远高于使用PCA降维结果。关键词：人脸识别；Gabor小波变换；LBP；距离测度西南科技大学本科生毕业论文3FacialfeatureextractionalgorithmresearchandsimulationbasedonGaborwaveletsAbstract：FaceRecognitionTechnology(FRT)isemergingasanactiveresearchareainthefieldofpatternrecognitionandartificialintelligence.Asabiometrictechnology,FRThasnumerousapplicationssuchasaccesscontrol,lawenforcement,e-commerce,videosurveillanceandsoon.Facefeatureextractionisthecoreofrecognitiontask,whichdirectlyimpactonclassificationvelocityandfacerecognitionability.Themaincontributionsofthisworkarelistedasfollows：(1)FacefeatureextractionalgorithmbasedonGabortransformisintroduced.ComparedwithFouriertransform，Gabortransformisprovedtobebetterinfacefeatureextraction.Andthen，AfastalgorithmofGaborTransformisintroduced.(2)AccordingtothedisadvantagesofGaborwavelet,ThispaperintroducedLBPoperator,theoperatorcanovercomethefaultsofGaborwavelettransformeffectively.Alsoitintroducedtheprincipleandrealizationprocesshowtoextractfeaturesfaceindetailed,andtheselectionofparametersofGaborfilteraredetailedinstructions.(3)FortheproblemtothefacefeaturevectorhighdimensioninGaborandLBPextraction,thispaperusedtheLPPandPCAdimensionreductionalgorithmreducedthedimension.(4)Inthefinalfacefeatureextraction,Bycalculatingthedistanceofthefeaturevectortolookthesimilaritybetweentheimagetoinquireandeachimageinimagedatabase.Thispartismainlytoverifytheextractedfacefeaturevector,andtocomparetwodifferentdimensionreductionmethod.Finally,Thispaperdothesimulationexperimentintheorlandyaletwofacedatabase.TheexperimentresultsshowthattheLPPdimensionreductionmethodisbetterthanPCAdimensionreductionmethod.Keywords:Facerecognition;Gaborwavelettransform;LBP;Distancemeasure西南科技大学本科生毕业论文4目录第1章绪论..........................................................61.1课题的背景和意义...............................................61.2当前人脸识别的国内外现状.......................................71.3课题的总设计思路...............................................81.4人脸特征提取的步骤及方法.......................................91.4.1人脸图像的预处理.........................................91.4.2基于Gabor小波算法进行人脸特征提取.......................91.4.3应用LBP算子............................................101.4.4采用PCA和LPP对提取的特征向量降维，以提高识别时间......101.4.5用距离测度法度量相似度进行结果测试......................101.5本章小结......................................................11第2章Gabor小波变换.................................................122.1引言..........................................................122.2Gabor小波变换的定义..........................................132.3Gabor变换在人脸识别中的应用..................................152.4Gabor滤波器..................................................152.4.1Gabor滤波器参数的选择...................................192.4.2Gabor滤波器的性质和计算方法.............................192.4.3Gabor变换的人脸特征提取.................................202.5Gabor快速算法图..............................................212.5.1快速傅里叶变换(FFT).....................................222.5.2Gabor变换的快速算法.....................................242.6本章小结.....................................25第3章LBP算法、PCA和LPP降维算法....................................263.1引言..........................................................263.2LBP基本算子..................................................263.2.1改进的LBP算子...........................................273.3PCA降维算法的实现原理........................................283.3.1PCA的基本概念...........................................29西南科技大学本科生毕业论文53.3.2PCA原理..................................................293.3.3PCA算法..................................................303.4LPP算子降维原理...............................................313.4.1LPP算法降维实现原理......................................313.5本章小结.......................................................32第四章距离测度法度量相似度比较......................................334.1引言...........................................................334.2距离测量简介...................................................334.3具体算法......................................................344.4ORL人脸库实验分析............................................354.4.1算法的比较..............................................364.5YALE人脸库实验...............................................374.6本章小结.......................................................38结论.................................................................39参考文献..............................................................40致谢...................................................