基于局部二元模式和韦伯局部描述符的人脸识别
|
摘要
众所周知,在计算机视觉和模式识别学科内,人脸识别是一个被研究很广泛的课题。如何找到一些行之有效的人脸特征来描述人脸是人脸识别技术的核心问题,现阶段成熟的人脸识别算法都是通过区分该特征来进行人脸识别的。一般情况下,从整体和局部的角度上来分的话,人脸识别算法可以分为基于全局的和基于局部的两类人脸识别算法。
全局方法如主成分分析(PCA)、核主成分分析(KPCA)和二维主成分分析(2DPCA)已经被广泛地研究。PCA是十分成熟的用于人脸特征提取的人脸识别算法之一,通过计算原输入的协方差矩阵的特征向量,PCA把高维的输入向量线性转换为低维的不相关向量。对于非线性PCA已经发展出了多种算法,核主成分分析(KPCA)就是其中一种,其在PCA中引入了核函数。2DPCA是基于二维图像矩阵的特征提取的一种人脸识别方法并取得了非常好的实验效果。它通过图像矩阵直接构造图像方差矩阵,并导出提取图像特征的特征向量。
局部方法中研究比较广泛的是LBP算子,灰度不变性和旋转不变性是LBP算子主要的特性也是主要的优点。尤其在最近几十年的时间内,LBP算子已经得到不断的发展和演化并广泛地应用于纹理分类、人脸识别等领域,但是LBP本身并不是完美的,也具有局限性,为了提高LBP的性能,本文进行了探索。
本文的主要贡献如下：
(1)对LBP算子的背景、基本原理和在人脸方面的应用进行了详细的梳理和总结,最后在试验结果上研究了基于LBP的人脸识别方法,通过试验验证了LBP方法的有效性,除此以外,还考察了各个参数对LBP方法的影响。
(2)将新出现的纹理识别算子WLD(韦伯局部描述符)运用于人脸识别。并在ORL、Yale和AR人脸库上考察了该算法的有效性,试验结果证明了该算法优于PCA、2DPCA和KPCA等主流人脸识别算法。
(3)受WLD的启发,将LBP和WLD进行结合提出了一种新的算子WLBP,它是对LBP算子的一种改进算法,通过在ORL、Yale和AR人脸库上的实验表明：WLBP识别效果优于现存的大部分人脸识别算法,诸如：PCA、KPCA、2DPCA、Gabor和LBP。
It is well known that face recognition is very popular in the research fields of pattern recognition and computer vision. Finding valid features to describe human faces is the key for face recognition. Generally, the current face recognition algorithms are classified into two categories, with one global-based and the other local-based.
The Principal Component Analysis (PCA), the Kernel Principal Component Analysis (KPCA) and the 2D Principal Component Analysis (2DPCA) from the global-based category, have been extensively studied. PCA is a famous method for feature extraction. PCA is capable of linearly converting a high dimensional input vector to a lower dimensional one by calculating the eigenvector of the initially input covariance matrix. Various nonlinear PCA methods have also been developed till now KPCA is one of them, which introduces the kernel function into the algorithm. Compared with the conventional PCA, the 2DPCA extracts features based on the two dimensional image matrix rather than the one dimensional matrix.2DPCA builds the image covariance matrix directly through the image matrix, and then exports the eigenvector of the image matrix.
The Local Binary Pattern (LBP) operator is one of the commonly used local-based face recognition method. With the properties of rotational invariance and gray invariance, the LBP operator has been developed constantly, and adopted in many areas, including texture classification, texture segmentation, and face recognition, to name a few. However, there is limitation of the LBP operator. This thesis aims at improving the performance of LBP. So far, I have done a lot to reach the aim.
The main contributions of this paper are summarized as follows:
(1) The background, fundamentals, and the application in face recognition of LBP were reviewed summarized in detail. At last, the method of LBP is proved effectual through the experiments in addition, I have examined the role of various parameters through all kind of experiments.
(2) The texture recognition operator, WLD (the Webber Local Descriptor), was successfully applied in face recognition, and satisfactory results were achieved. The experiments conducted on ORL, Yale and AR human face database, showed that WLD is superior to most existing face recognition methods, such as PCA, KPCA,and2DPCA.
(3) With the inspiration of WLD, LBP and WLD were combined to come up with a new operator, named WLBP, which improves the LBP operator. The experiments conducted on ORL, Yale and AR human face database, showed that WLBP is superior to most existing face recognition methods, such as PCA, KPCA,2DPCA,Gabor and LBP.
全局方法如主成分分析(PCA)、核主成分分析(KPCA)和二维主成分分析(2DPCA)已经被广泛地研究。PCA是十分成熟的用于人脸特征提取的人脸识别算法之一,通过计算原输入的协方差矩阵的特征向量,PCA把高维的输入向量线性转换为低维的不相关向量。对于非线性PCA已经发展出了多种算法,核主成分分析(KPCA)就是其中一种,其在PCA中引入了核函数。2DPCA是基于二维图像矩阵的特征提取的一种人脸识别方法并取得了非常好的实验效果。它通过图像矩阵直接构造图像方差矩阵,并导出提取图像特征的特征向量。
局部方法中研究比较广泛的是LBP算子,灰度不变性和旋转不变性是LBP算子主要的特性也是主要的优点。尤其在最近几十年的时间内,LBP算子已经得到不断的发展和演化并广泛地应用于纹理分类、人脸识别等领域,但是LBP本身并不是完美的,也具有局限性,为了提高LBP的性能,本文进行了探索。
本文的主要贡献如下：
(1)对LBP算子的背景、基本原理和在人脸方面的应用进行了详细的梳理和总结,最后在试验结果上研究了基于LBP的人脸识别方法,通过试验验证了LBP方法的有效性,除此以外,还考察了各个参数对LBP方法的影响。
(2)将新出现的纹理识别算子WLD(韦伯局部描述符)运用于人脸识别。并在ORL、Yale和AR人脸库上考察了该算法的有效性,试验结果证明了该算法优于PCA、2DPCA和KPCA等主流人脸识别算法。
(3)受WLD的启发,将LBP和WLD进行结合提出了一种新的算子WLBP,它是对LBP算子的一种改进算法,通过在ORL、Yale和AR人脸库上的实验表明：WLBP识别效果优于现存的大部分人脸识别算法,诸如：PCA、KPCA、2DPCA、Gabor和LBP。
It is well known that face recognition is very popular in the research fields of pattern recognition and computer vision. Finding valid features to describe human faces is the key for face recognition. Generally, the current face recognition algorithms are classified into two categories, with one global-based and the other local-based.
The Principal Component Analysis (PCA), the Kernel Principal Component Analysis (KPCA) and the 2D Principal Component Analysis (2DPCA) from the global-based category, have been extensively studied. PCA is a famous method for feature extraction. PCA is capable of linearly converting a high dimensional input vector to a lower dimensional one by calculating the eigenvector of the initially input covariance matrix. Various nonlinear PCA methods have also been developed till now KPCA is one of them, which introduces the kernel function into the algorithm. Compared with the conventional PCA, the 2DPCA extracts features based on the two dimensional image matrix rather than the one dimensional matrix.2DPCA builds the image covariance matrix directly through the image matrix, and then exports the eigenvector of the image matrix.
The Local Binary Pattern (LBP) operator is one of the commonly used local-based face recognition method. With the properties of rotational invariance and gray invariance, the LBP operator has been developed constantly, and adopted in many areas, including texture classification, texture segmentation, and face recognition, to name a few. However, there is limitation of the LBP operator. This thesis aims at improving the performance of LBP. So far, I have done a lot to reach the aim.
The main contributions of this paper are summarized as follows:
(1) The background, fundamentals, and the application in face recognition of LBP were reviewed summarized in detail. At last, the method of LBP is proved effectual through the experiments in addition, I have examined the role of various parameters through all kind of experiments.
(2) The texture recognition operator, WLD (the Webber Local Descriptor), was successfully applied in face recognition, and satisfactory results were achieved. The experiments conducted on ORL, Yale and AR human face database, showed that WLD is superior to most existing face recognition methods, such as PCA, KPCA,and2DPCA.
(3) With the inspiration of WLD, LBP and WLD were combined to come up with a new operator, named WLBP, which improves the LBP operator. The experiments conducted on ORL, Yale and AR human face database, showed that WLBP is superior to most existing face recognition methods, such as PCA, KPCA,2DPCA,Gabor and LBP.
引文
[1]Ratha N K, Senior A, Boile R M. Automated biometrics.In:Singh S, Murshed N, KropatschW,eds.AdvancesinPatternRecognition-ICAPP2001.Berlin:Springer-verlag,2001, 2013:445-453.
[2]Blometrics Market and Industry Report 2007-2012[R]. Internation Blometric Group, 2007.
[3]A.K.Jain,A.Ross,S.Prabhakar,"An Introduction to Biometric Recognition",IEEE Trans.on Circuits and Systems for Video Technology,Vol.14,No.1,pp 4-19,Janua 2004
[4]Zhaow, Cheil, Appar, etal. Facerecognition:A literature survey [J]. ACM Computing Surveys,2003,35(4):399,458.
[5]Chellappa R, Wilson CL, Sirohey S. Human and machine recognition of faces:a survey. Proc.IEEE.1995,83(5):705-740.
[6]山世光.人脸识别中若干关键问题的研究[D].中国科学院研究生院博士学位论文,2004.
[7]Chan H.and Bledsoe W.W., Aman-machine facial recognition system:some Prelimin-ary result, Techinical report, Panoramic Reserch Inc.Cal,1965.
[8]PhilliP P.J., MoonH., etc. The FERET Evaluation Methodology for Face Recogniti-on Algorithms, IEEE Transactions on PAMI,2000,22(10),1090-1104.
[9]A.S.Georghiades, D.J.Kriegman, P.N.Bellhumeur. Illumination Cones For Recognitio-n Under Variable Lighting Face. Proc.of IEEE CVPR.1998,52-58.
[10]Volker Blanz Thomas Vetter,A Morphable Model For the Synthesis of 3D Faces, SIG GRAPH99.1999.
[11]J. Yang, D. Zhang, A. F. Frangi, and J. Yang, Two-dimensional PCA:A new app-roach to appearance-based face representation and recognition, IEEE Trans. Patter-nanal.Machine Intell.26 (1) (2004) 131-137.
[12]张文超等基于局部Gabor变化直方图序列的人脸描述与识别Journal of Software, Vol.17, No.12, December 2006, pp.2508-2517
[13]Timo Ahonen, Abdenour Hadid, and Matti Pietikainen Face Recognition with Local Binary Patterns ECCA 2004, LNCS 3021, pp.469-481,2004.
[14]孔万增,基于学习的人脸识别研究,浙江大学博士学位论文,2008.
[15]吴巾一,周德龙,人脸识别方法综述,计算机应用研究,Vol.26,No.9,Sep.2009
[16]H.Chan and W.W.Bledsoe.A man-machine facial recognition system:some prelimin- ary results[R].Technical report,Panoramic Research Inc.,Cal,1965.
[17]M.D.Kelly,Visual identification of people by computer[R]. Technical Report.AI-130, Stanford AI Project,Stanford,CA,1970.
[18]R.Brunelli,T.Poggio.Feature recognition:features versus templates[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,
[19]Jain A, Huang J.Intergrating independent components andl inear Discriminant ana-lysis for gender classifieation[A]. Automatiec Face and Gesture Recognition[C].200 4,159-163.
[20]M.Lades,J.C.Vorbruggen,J.Buhmann,J.Lange,C.Von der Malsburg,R.P.Wurtz,et al.Dist ortion invariant object recognition in the dynamic link architecture[J].IEEE Trans, on Computers,1993,42(3):300-311.
[21]周凯,基于局部二值模式的人脸识别方法研究,中南大学博士论文,2009
[22]Laurenz Wiskott,Jean-Marc Fellous,Norbert Kriiger,Christoph von der Malsburg.Face recognition by elastic bunch graph matching[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(7):775-779.
[23]Ojala T,Pietikainen M,Maenpaa T.Multiresolution gray-scale and rotation Invariant texture classification with local binary patterns.IEEE Trans on PAML,2002,24(7): 971=987.
[24]张文超,山世光,张洪明等.基于局部Gabor变化直方图序列的人脸描述与识别[J].软件学报,2006,17(12)：2508-2517
[25]Peter N.Belhumeur,Joao P.Hespanha,David J.Kriegman.Eigenfaces vs.Fisherfaces: recognition using class specific linear projection[J]. IEEE Transactions on Pattern Analysis andMachine Intelligence,1997,19(7):711-720.
[26]T.Kohonen.Self-organization and associative memory[M].New York:Springer-Verlag, 1988.
[27]刘永信等,人脸识别方法综述,内蒙古大学学报,Jul.2009,Vol.40 No.4
[28]Vapnik V.The Nature of Statistical Learning Theory.Berlin:Springer Verlag,1995
[29]SAMARIA F,YOUNG S. HMM based architecture for face identification[j].ImageAnd Computer Vision,1994,12(8):537-54
[30]NEFIANAV, MONSONHH. Hidden Markov models forfac erecogniton[C]//Proc of IEEE International Conference on Acoustics, Speech and sing Processing.1998:141-145.
[31]刘小军,王东峰,张丽飞,等.一种基于奇异值分解和隐马尔可夫模型的人脸识别 方法[J].计算机学报,2003,26(3)：340.344.
[32]Li S Z,Jain A K.Introduction In:Hand book of face recognition.Editor by Li S Z,Jain A K.Springer Scence Business Media,Inc.2005.
[33]Philllips P J,Flynn P J,Scruggs T,et al.Overview of the face recognition and challenge In Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition,San Diego,2005,1:947-954.
[34]付晓峰.基于二元模式的人脸识别与表情识别研究[D].浙江大学博士毕业论文2008
[35]金一.人脸识别中的若干算法研究[D],北京交通大学博士论文,2009
[36]马莉,范影乐.纹理图像分析[M].北京：科学出版社,2009
[37]刘小民.纹理研究及其应用综述[J].测控技术,2008,27(5)：4-9
[38]刘小民.纹理研究综述[J].计算机应用研究,2008,25(8)：2284-2288
[39]Ojala T, Pietikainen M, Harwood D. A comparative study of texture measures with classification based on featured distributions.1996
[40]钱艺,童卫青,王基帆.基于多尺度LBP的人脸识别算法.图形图像.2009.8
[41]S.Kullback. Information theory and statistics [M]. Dover Publications, New York, 1969.
[42]R.Sokal, F.Rohlf. Biometry [M]. Freeman:San Francisco,1969.
[43]A.M. Martinez, R. Benavente. "The AR Face Database". CVC Technical Report, no.24, June 1998.
[44]A.M.Martinez, R.Benavente, "The AR Face Database", http://rvll.ecn.purdue.edu/-a leix/aleix face DB.html,2003.
[45]Jie Chen, Member, Shiguang Shan, etal. WLD:A Robust Local Image Descriptor, IEEE transactions on pattern analysis and machine intelligence,2009.
[46]Jianhong Shen, Weber's Law and Weberized TV Restoratio Physica D:Nonlinear Phenomena,2003,175:241-251.
[47]Stephen DelMarco, Victor Tom, Helen Webb. A Theory of Automatic Parameter Selection for Feature Extraction With Application to Feature-Based Multisensor Image Registration. IEEE transactions on image processing.2007,16(11):2733-2742.
[2]Blometrics Market and Industry Report 2007-2012[R]. Internation Blometric Group, 2007.
[3]A.K.Jain,A.Ross,S.Prabhakar,"An Introduction to Biometric Recognition",IEEE Trans.on Circuits and Systems for Video Technology,Vol.14,No.1,pp 4-19,Janua 2004
[4]Zhaow, Cheil, Appar, etal. Facerecognition:A literature survey [J]. ACM Computing Surveys,2003,35(4):399,458.
[5]Chellappa R, Wilson CL, Sirohey S. Human and machine recognition of faces:a survey. Proc.IEEE.1995,83(5):705-740.
[6]山世光.人脸识别中若干关键问题的研究[D].中国科学院研究生院博士学位论文,2004.
[7]Chan H.and Bledsoe W.W., Aman-machine facial recognition system:some Prelimin-ary result, Techinical report, Panoramic Reserch Inc.Cal,1965.
[8]PhilliP P.J., MoonH., etc. The FERET Evaluation Methodology for Face Recogniti-on Algorithms, IEEE Transactions on PAMI,2000,22(10),1090-1104.
[9]A.S.Georghiades, D.J.Kriegman, P.N.Bellhumeur. Illumination Cones For Recognitio-n Under Variable Lighting Face. Proc.of IEEE CVPR.1998,52-58.
[10]Volker Blanz Thomas Vetter,A Morphable Model For the Synthesis of 3D Faces, SIG GRAPH99.1999.
[11]J. Yang, D. Zhang, A. F. Frangi, and J. Yang, Two-dimensional PCA:A new app-roach to appearance-based face representation and recognition, IEEE Trans. Patter-nanal.Machine Intell.26 (1) (2004) 131-137.
[12]张文超等基于局部Gabor变化直方图序列的人脸描述与识别Journal of Software, Vol.17, No.12, December 2006, pp.2508-2517
[13]Timo Ahonen, Abdenour Hadid, and Matti Pietikainen Face Recognition with Local Binary Patterns ECCA 2004, LNCS 3021, pp.469-481,2004.
[14]孔万增,基于学习的人脸识别研究,浙江大学博士学位论文,2008.
[15]吴巾一,周德龙,人脸识别方法综述,计算机应用研究,Vol.26,No.9,Sep.2009
[16]H.Chan and W.W.Bledsoe.A man-machine facial recognition system:some prelimin- ary results[R].Technical report,Panoramic Research Inc.,Cal,1965.
[17]M.D.Kelly,Visual identification of people by computer[R]. Technical Report.AI-130, Stanford AI Project,Stanford,CA,1970.
[18]R.Brunelli,T.Poggio.Feature recognition:features versus templates[J].IEEE Trans.on Pattern Analysis and Machine Intelligence,
[19]Jain A, Huang J.Intergrating independent components andl inear Discriminant ana-lysis for gender classifieation[A]. Automatiec Face and Gesture Recognition[C].200 4,159-163.
[20]M.Lades,J.C.Vorbruggen,J.Buhmann,J.Lange,C.Von der Malsburg,R.P.Wurtz,et al.Dist ortion invariant object recognition in the dynamic link architecture[J].IEEE Trans, on Computers,1993,42(3):300-311.
[21]周凯,基于局部二值模式的人脸识别方法研究,中南大学博士论文,2009
[22]Laurenz Wiskott,Jean-Marc Fellous,Norbert Kriiger,Christoph von der Malsburg.Face recognition by elastic bunch graph matching[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,1997,19(7):775-779.
[23]Ojala T,Pietikainen M,Maenpaa T.Multiresolution gray-scale and rotation Invariant texture classification with local binary patterns.IEEE Trans on PAML,2002,24(7): 971=987.
[24]张文超,山世光,张洪明等.基于局部Gabor变化直方图序列的人脸描述与识别[J].软件学报,2006,17(12)：2508-2517
[25]Peter N.Belhumeur,Joao P.Hespanha,David J.Kriegman.Eigenfaces vs.Fisherfaces: recognition using class specific linear projection[J]. IEEE Transactions on Pattern Analysis andMachine Intelligence,1997,19(7):711-720.
[26]T.Kohonen.Self-organization and associative memory[M].New York:Springer-Verlag, 1988.
[27]刘永信等,人脸识别方法综述,内蒙古大学学报,Jul.2009,Vol.40 No.4
[28]Vapnik V.The Nature of Statistical Learning Theory.Berlin:Springer Verlag,1995
[29]SAMARIA F,YOUNG S. HMM based architecture for face identification[j].ImageAnd Computer Vision,1994,12(8):537-54
[30]NEFIANAV, MONSONHH. Hidden Markov models forfac erecogniton[C]//Proc of IEEE International Conference on Acoustics, Speech and sing Processing.1998:141-145.
[31]刘小军,王东峰,张丽飞,等.一种基于奇异值分解和隐马尔可夫模型的人脸识别 方法[J].计算机学报,2003,26(3)：340.344.
[32]Li S Z,Jain A K.Introduction In:Hand book of face recognition.Editor by Li S Z,Jain A K.Springer Scence Business Media,Inc.2005.
[33]Philllips P J,Flynn P J,Scruggs T,et al.Overview of the face recognition and challenge In Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition,San Diego,2005,1:947-954.
[34]付晓峰.基于二元模式的人脸识别与表情识别研究[D].浙江大学博士毕业论文2008
[35]金一.人脸识别中的若干算法研究[D],北京交通大学博士论文,2009
[36]马莉,范影乐.纹理图像分析[M].北京：科学出版社,2009
[37]刘小民.纹理研究及其应用综述[J].测控技术,2008,27(5)：4-9
[38]刘小民.纹理研究综述[J].计算机应用研究,2008,25(8)：2284-2288
[39]Ojala T, Pietikainen M, Harwood D. A comparative study of texture measures with classification based on featured distributions.1996
[40]钱艺,童卫青,王基帆.基于多尺度LBP的人脸识别算法.图形图像.2009.8
[41]S.Kullback. Information theory and statistics [M]. Dover Publications, New York, 1969.
[42]R.Sokal, F.Rohlf. Biometry [M]. Freeman:San Francisco,1969.
[43]A.M. Martinez, R. Benavente. "The AR Face Database". CVC Technical Report, no.24, June 1998.
[44]A.M.Martinez, R.Benavente, "The AR Face Database", http://rvll.ecn.purdue.edu/-a leix/aleix face DB.html,2003.
[45]Jie Chen, Member, Shiguang Shan, etal. WLD:A Robust Local Image Descriptor, IEEE transactions on pattern analysis and machine intelligence,2009.
[46]Jianhong Shen, Weber's Law and Weberized TV Restoratio Physica D:Nonlinear Phenomena,2003,175:241-251.
[47]Stephen DelMarco, Victor Tom, Helen Webb. A Theory of Automatic Parameter Selection for Feature Extraction With Application to Feature-Based Multisensor Image Registration. IEEE transactions on image processing.2007,16(11):2733-2742.