多视角下多模板人脸特征定位方法
|
摘要
为了提高多视角下人脸ASM算法的定位精度,提出一种使用Candide-3模型估计人脸视角,使用全局模板和局部模板相结合的方法定位人脸特征点的多视角人脸ASM方法。建立不同视角下的形状统计模型,从中选择若干关键特征点,用SVM分类器建立纹理模型;在特征点定位中通过检测出的眼角、嘴角、鼻下点用Candide-3估计人脸姿态,选择相应视角的全局模板总体定位,利用局部模板估计未检测到的特征点,再用全局模板校正,得到人脸特征点位置信息。实验结果表明,该方法显著改善了多视角下人脸特征点定位不准确的问题,有效提高了各特征点的定位精度。
To improve the active shape model(ASM)location accuracy on multi-view face,the multi-view ASM based algorithm which estimates view by Candide-3and combines the global template and the local templates to obtain the feature points is proposed.First,the statistical shape models under different view are built.Some key feature points are selected and their texture models are built by SVM.In the subsequent searching,the face pose is estimated by Candide-3based on the subnasal points and the corners of eyes and mouth,then the global model under the estimated view is selected to perform a coarse location,then the local templates are used to estimate the points undetected out,and finally the global model is used to check up the shape to obtain the feature points information.Experimental results show that the proposed method significantly improves the location accuracy of the traditional ASM on multi-view face,and enhances the detection accuracy of all the feature points.
To improve the active shape model(ASM)location accuracy on multi-view face,the multi-view ASM based algorithm which estimates view by Candide-3and combines the global template and the local templates to obtain the feature points is proposed.First,the statistical shape models under different view are built.Some key feature points are selected and their texture models are built by SVM.In the subsequent searching,the face pose is estimated by Candide-3based on the subnasal points and the corners of eyes and mouth,then the global model under the estimated view is selected to perform a coarse location,then the local templates are used to estimate the points undetected out,and finally the global model is used to check up the shape to obtain the feature points information.Experimental results show that the proposed method significantly improves the location accuracy of the traditional ASM on multi-view face,and enhances the detection accuracy of all the feature points.
引文
[1]Cootes T F,Taylor C J,Cooper D H,et al.Active shape models-their training and their application[J].Computer Vision and Image Understanding,1995,61(1):38-59.
[2]ZHU Jie,TANG Zhenmin.ASM and color Gabor features for facial feature extraction[J].Computer Science,2010,37(4):265-268(in Chinese).[朱杰,唐振民.ASM与彩色Gabor特征相结合的人脸关键特征点提取[J].计算机科学,2010,37(4):265-268.]
[3]NIU Zhiheng.Study on facial landmark localization[D].Harbin:Harbin Institute of Technology,2009:37-51(in Chinese).[牛志恒.面部特征点定位方法研究[D].哈尔滨:哈尔滨工业大学,2009:37-51.]
[4]LI Hao,XIE Chen,TANG Chaojing.An improved algorithm for facial feature location by multi-template ASM[J].Journal of Computer Aided Design&Computer Graphics,2010,22(10):1762-1768(in Chinese).[李皓,谢琛,唐朝京.改进的多模板ASM人脸面部特征定位算法[J].计算机辅助设计与图形学学报,2010,22(10):1762-1768.]
[5]WANG Liting,DING Xiaoqing,FANG Chi.A novel method for robust and automatic facial features localization[J].Acta Automatica Sinica,2009,35(1):9-16(in Chinese).[王丽婷,丁晓青,方驰.一种鲁棒的全自动人脸特征点定位方法[J].自动化学报,2009,35(1):9-16.]
[6]LI Weisheng,ZHANG Yan,ZHOU Lifang.ASM facial feature location under attitude changes[J].Application Research of Computers,2012,29(12):4764-4766(in Chinese).[李伟生,张燕,周丽芳.一种姿态变化下的ASM人脸特征定位方法[J].计算机应用研究,2012,29(12):4764-4766.]
[7]FAN Xiaojiu,PENG Qiang,Jim X Chen.Rapid and automatic3Dface modeling using active appearance models[J].Journal of Computer Aided Design&Computer Graphics,2009,21(3):668-673(in Chinese).[范小九,彭强,Jim X Chen.利用主动外观模型进行3D人脸的快速自动建模[J].计算机辅助设计与图形学学报,2009,21(3):668-673.]
[8]Fu Youjia,Li Jianwei,Xiang Ruxi.Enhance ASM based on DCT-SVM for facial feature points localization[C]//Taiwan:Proceeding of the 2nd International Conference on Frontiers of Manufacturing and Design Science,2011:820-824.
[9]Fu Youjia,He Yan,Li Jianwei,et al.Robust facial features localization on rotation arbitrary multi-view face in complex background[J].Journal of Computers,2011,6(2):337-342.
[10]Ahlberg J.Candide-3an updated parameterized face[R].Sweden:Linoping University,2001.
[2]ZHU Jie,TANG Zhenmin.ASM and color Gabor features for facial feature extraction[J].Computer Science,2010,37(4):265-268(in Chinese).[朱杰,唐振民.ASM与彩色Gabor特征相结合的人脸关键特征点提取[J].计算机科学,2010,37(4):265-268.]
[3]NIU Zhiheng.Study on facial landmark localization[D].Harbin:Harbin Institute of Technology,2009:37-51(in Chinese).[牛志恒.面部特征点定位方法研究[D].哈尔滨:哈尔滨工业大学,2009:37-51.]
[4]LI Hao,XIE Chen,TANG Chaojing.An improved algorithm for facial feature location by multi-template ASM[J].Journal of Computer Aided Design&Computer Graphics,2010,22(10):1762-1768(in Chinese).[李皓,谢琛,唐朝京.改进的多模板ASM人脸面部特征定位算法[J].计算机辅助设计与图形学学报,2010,22(10):1762-1768.]
[5]WANG Liting,DING Xiaoqing,FANG Chi.A novel method for robust and automatic facial features localization[J].Acta Automatica Sinica,2009,35(1):9-16(in Chinese).[王丽婷,丁晓青,方驰.一种鲁棒的全自动人脸特征点定位方法[J].自动化学报,2009,35(1):9-16.]
[6]LI Weisheng,ZHANG Yan,ZHOU Lifang.ASM facial feature location under attitude changes[J].Application Research of Computers,2012,29(12):4764-4766(in Chinese).[李伟生,张燕,周丽芳.一种姿态变化下的ASM人脸特征定位方法[J].计算机应用研究,2012,29(12):4764-4766.]
[7]FAN Xiaojiu,PENG Qiang,Jim X Chen.Rapid and automatic3Dface modeling using active appearance models[J].Journal of Computer Aided Design&Computer Graphics,2009,21(3):668-673(in Chinese).[范小九,彭强,Jim X Chen.利用主动外观模型进行3D人脸的快速自动建模[J].计算机辅助设计与图形学学报,2009,21(3):668-673.]
[8]Fu Youjia,Li Jianwei,Xiang Ruxi.Enhance ASM based on DCT-SVM for facial feature points localization[C]//Taiwan:Proceeding of the 2nd International Conference on Frontiers of Manufacturing and Design Science,2011:820-824.
[9]Fu Youjia,He Yan,Li Jianwei,et al.Robust facial features localization on rotation arbitrary multi-view face in complex background[J].Journal of Computers,2011,6(2):337-342.
[10]Ahlberg J.Candide-3an updated parameterized face[R].Sweden:Linoping University,2001.