人脸特征点定位算法研究及其在人脸卡通肖像中的应用
|
摘要
卡通人脸是指既具夸张风格,又不失真实人脸的可识别特征的艺术形象。近年来,随着计算机技术的飞速发展,如何利用计算机自动生成卡通人脸肖像已经成为国内外的研究热点之一。本文对人脸卡通肖像绘制方法从特征定位、图像变形以及肖像提取三个方面展开研究。
首先：本文参考MPEG-4标准,建立了一个由78个特征点表示的人脸模型,并采用主动形状模型(Active Shape Model,ASM)方法进行人脸特征定位。然而,该方法易受初始情况、光照等因素的影响,会因模型畸变导致特征点不能提取准确。为此,本文提出了一种根据特征点间的位置关系,建立关联度评价函数,由此修正特征点搜索错误。实验结果表明,与传统ASM相比,本算法可获得高准确度的定位效果。
其次：在夸张人脸个性特征时,只需要通过用户交互,选取需要变形的人脸特征,就能够对该特征对应的特征点的位置进行调整。再利用基于特征线对的图像变形方法,可实现对人脸图像进行变形,获得具有夸张效果的人脸图像。
最后：在肖像化的过程中,本文利用特征点的位置,并定义了一定的笔触,使得人脸轮廓线会产生渐变的效果,使获得的肖像画更具立体特征,更生动。
实验结果表明,本系统可准确提出人脸的特征点,并且可以生动自然生成人脸卡通肖像画。
Caricature Face is an artistic image which refers to be both exaggerated and without losing the true face's identifiable features. In recent years, with the rapid development of computer technology, the issue of how to use computer to automatically generate caricature face portrait has become a research hot spot at home and abroad. In this paper, the research on face caricature portrait is studied from facial feature location, image warp, as well as face portrait three aspects.
First of all, according to MPEG-4 standard, this paper use 78 feature points to express human face model, and use active shape model (ASM) method to locate feature points. However, due to lighting condition and other factors, it will lead to inaccurate feature point location due to model distortion. So, this paper proposes to construct an association degree evaluation function according to relations of feature points. This revises error feature points. The experimental results show that the algorithm is able to obtain highly accurate result than the traditional ASM.
Second:When exaggerating facial personal features, just need user interaction, select the required deformation facial features, then can correspond to the characteristics of the location of feature points to adjust, and utilize image distortion algorithm based on feature lines can realized image deformation and obtain an exaggerated effects of human face images.
In the end:In the process of portrait, this paper uses the positions of feature points and defines a brush stroke which can get gradient facial contours. So, the portrait can be more and more vivid.
The experimental results show that the system can make an accurate facial feature points location, and can make a lively face cartoon portraits.
首先：本文参考MPEG-4标准,建立了一个由78个特征点表示的人脸模型,并采用主动形状模型(Active Shape Model,ASM)方法进行人脸特征定位。然而,该方法易受初始情况、光照等因素的影响,会因模型畸变导致特征点不能提取准确。为此,本文提出了一种根据特征点间的位置关系,建立关联度评价函数,由此修正特征点搜索错误。实验结果表明,与传统ASM相比,本算法可获得高准确度的定位效果。
其次：在夸张人脸个性特征时,只需要通过用户交互,选取需要变形的人脸特征,就能够对该特征对应的特征点的位置进行调整。再利用基于特征线对的图像变形方法,可实现对人脸图像进行变形,获得具有夸张效果的人脸图像。
最后：在肖像化的过程中,本文利用特征点的位置,并定义了一定的笔触,使得人脸轮廓线会产生渐变的效果,使获得的肖像画更具立体特征,更生动。
实验结果表明,本系统可准确提出人脸的特征点,并且可以生动自然生成人脸卡通肖像画。
Caricature Face is an artistic image which refers to be both exaggerated and without losing the true face's identifiable features. In recent years, with the rapid development of computer technology, the issue of how to use computer to automatically generate caricature face portrait has become a research hot spot at home and abroad. In this paper, the research on face caricature portrait is studied from facial feature location, image warp, as well as face portrait three aspects.
First of all, according to MPEG-4 standard, this paper use 78 feature points to express human face model, and use active shape model (ASM) method to locate feature points. However, due to lighting condition and other factors, it will lead to inaccurate feature point location due to model distortion. So, this paper proposes to construct an association degree evaluation function according to relations of feature points. This revises error feature points. The experimental results show that the algorithm is able to obtain highly accurate result than the traditional ASM.
Second:When exaggerating facial personal features, just need user interaction, select the required deformation facial features, then can correspond to the characteristics of the location of feature points to adjust, and utilize image distortion algorithm based on feature lines can realized image deformation and obtain an exaggerated effects of human face images.
In the end:In the process of portrait, this paper uses the positions of feature points and defines a brush stroke which can get gradient facial contours. So, the portrait can be more and more vivid.
The experimental results show that the system can make an accurate facial feature points location, and can make a lively face cartoon portraits.
引文
[1]倪菲,付庄,曹其新等.肖像漫画绘制机器人技术研究.Chinese Journal of Nature[J],2007.4, Vol.29 No.4.
[2]沈向洋,徐迎庆.数字化动漫：计算机自动绘制人脸卡通画像.前沿科学[J],2007,Vol.24(6)：201-203.
[3]董致成,杨志耀,白敦文.二维卡通脸谱产生系统(Ⅱ)-以虚拟脸谱特征向量为基础之架构：[硕士学位论文][D].台湾：国立台湾海洋大学,2007.
[4]周仁琴,周经野,陈益强等.基于特征发现的卡通人脸肖像生成.计算机辅助设计与图形学学报[J],2006,144(15)：71～72.
[5]阎芳,费广正,柳婷婷等.漫画风格的人脸卡通生成算法.计算机辅助设计与图形学学报[J],2007,19(4).
[6]李龙生,周经野,陈益强等.面向移动的卡通人脸生成系统.计算机科学[J],2007,Vol.34 No.1.
[7]万里红,董兰芳,沈育平等.基于特征网格的面部图像变形算法.中国图象图形学报[J],2002,Vol 7(A),No.3.
[8]张静波,张桂林等.基于样本学习的人像线条画生成系统.软件学报[J].2003,14(2).
[9]胡玲.基于样本学习的人脸个性化卡通模型系统性能分析与优化设计：[硕士学位论文][D].南京：东南大学,2005.
[10]江佩颖,李蔡彦等.以脸部特征为基础之夸张肖像画产生系统：[硕士学位论文][D].台湾：国立台湾海洋大学,2007.
[11]陈洪,郑南宁,梁林等.基于样本学习的肖像画自动生成算法系统.计算机学报[J].
[12]万里红,董兰芳,沈育平,陈意云.人物肖像画及其卡通动画仿真系统的建立.计算机仿真[J],2008,44(15)：71～72.
[13]张翼,陈纯.基于视觉特征提取的表情人脸动画生成：[硕士学位论文][D].杭州：浙江大学,2008.
[14]刘振安.基于SVG的卡通人脸图形自动生成法.测控技术[J],2006,vol(5)：101-107.
[15]林维训,潘纲,吴朝晖,潘云鹤.脸部特征定位方法综述.中国图象图形学报[J],2006.5,vol(5)：101-107.
[16]吴证.基于样本学习的人像线条画生成系统：[硕士学位论文][D].上海：上海交通大 学,2007.
[17]范玉华.基于ASM的人脸面部关键特征点定位算法研究：[硕士学位论文][D].郑州：河南科技大学,2008.
[18]曾阳.人脸特征提取方法研究：[硕士学位论文][D].长沙：中南大学,2008.
[19]吕宝华.人脸特征定位方法研究：[硕士学位论文][D].西安：西北工业大学大学,2007.
[20]吴证.人脸特征点定位及其应用：[硕士学位论文][D].上海：上海交通大学,2007.
[21]Cootes T F. Locating Faces Using Statistical Feature Detectors[C]. In:Proceedings of 2nd International Conference on Automatic Face and Gesture Recognition, Killington, 1996,204 —209.
[22]Thaddeus Beier. Feature-based Image Metamorphosis. Computer Graphics[J].1992, Vol.75:3-24.
[23]HER TZMANN A. Painterly Rendering With Curved Brush Strokes of Multiple Sizes[C]. In:Proceedings of the 25th Annual Conference on Computer Graphics and Interactive Techniques,1998.
[24]Ergun Akleman. Making Extreme Caricatures with a New Interactive 2D Deformation Techique With Simplicial. Complexes. Computer Graphics[J].1992, Vol.75:3-24
[25]Son Lam Phung, Douglas Chai, Bouzerdoum. Skin Color Based Faee Detection[C].In:Proeeeding of Intelligent Information Systems Conference.Australian and New Zealand,2001,171—176.
[26]HeiseleB, HoP, Poggio T. Face recognition with support veetor machines:global versus component-based approach[C].In:Proeeeding of International Conference on Computer Vision. Vancouver,Canada,July,2001,2,688—694.
[27]Yang G, Huang T S. Human face detection in a complex background[J]. Pattern Recognition,1994,27 (1):53—63.
[28]Feng G C, Yuen P C. Variance Projection function and its application to eye detection for human face recognition[J]. Pattern Recognition Letters,1998,19 (9):899—906.
[29]Yuille A L, Cohen D S, Halinan P W. Feature extraction from faces using deformable templates[C].In:Proeeedings of IEEE Computer Soc. Conf. on Computer Vision and Pattern Recognition. Washington,DC,USA,1989,104—109.
[30]Cootes T F, Taylor C J, CooPer D H,et al. Active shape models-their training and application[J].Computer Vision and Image Understanding,1995,61 (1):38—59.
[31]Hsu R L, Mottaleb M A, Jain A K. Face detection in color Images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,24 (5):696—706.
[32]Waite J, Vincent J. A Probabilistic framework for neural network facial feature location[J].British Telecom Technology Journal,1992,10 (3):20—29.
[33]Turk M, Pentland A. Eigenfaces for recognition[J]. Journal of cognitive Neuroscience,1991,3 (1):71—86.
[34]Pan G, Lin W X, Wu Z H,et al. An eye detcetion system based on SVM filter[C].In:Proeeedings of Electronic Imaging and Multimedia Technology. Shanghai,China,2002,326 — 331.
[35]PJ.PhilliPs,H.Moon,etc. "The FERET Evaluation Methodology of Face-Reeongition Algorithms',IEEE Transactions on PAMI,VOL.22,No.10,pp 1090-1104,2000.
[36]BeierT, Neely S. Feature-based image metamorphosis[A].Poreeedings of ACM SIGGRAPH'92[C].Chicago:ACM Press,1992.35—42.
[37]RuPercht D,Mller H. Image warping with scattered data interpolation[J].IEEE Computer Graph Appl,1995,15:37—43.
[38]Hassnaien A E, Nakajima M. Image morphing of the facial image transformation based on Navier elastic body splines. Geneva,Switzerland,IEEE Computer Society Press,Los Alamitos,CA,pp 119—125.
[39]LeeS,ChwaK-Y,Hahn J, Shin SY (1996).Image morphing using deformation techniques.J Visualization Comput Anim 7:3-23.
[40]Gao P, Sederberg T W. A work minimization approach to imagemorphing[J].Visual ComPuter,1998.
[41]Michael Kass, Andrew Wiktin, and Demetri Terzopoulos. Snkaes:Active contour models. Intermational Joural of Computer Vision[J],pages 321—331.
[2]沈向洋,徐迎庆.数字化动漫：计算机自动绘制人脸卡通画像.前沿科学[J],2007,Vol.24(6)：201-203.
[3]董致成,杨志耀,白敦文.二维卡通脸谱产生系统(Ⅱ)-以虚拟脸谱特征向量为基础之架构：[硕士学位论文][D].台湾：国立台湾海洋大学,2007.
[4]周仁琴,周经野,陈益强等.基于特征发现的卡通人脸肖像生成.计算机辅助设计与图形学学报[J],2006,144(15)：71～72.
[5]阎芳,费广正,柳婷婷等.漫画风格的人脸卡通生成算法.计算机辅助设计与图形学学报[J],2007,19(4).
[6]李龙生,周经野,陈益强等.面向移动的卡通人脸生成系统.计算机科学[J],2007,Vol.34 No.1.
[7]万里红,董兰芳,沈育平等.基于特征网格的面部图像变形算法.中国图象图形学报[J],2002,Vol 7(A),No.3.
[8]张静波,张桂林等.基于样本学习的人像线条画生成系统.软件学报[J].2003,14(2).
[9]胡玲.基于样本学习的人脸个性化卡通模型系统性能分析与优化设计：[硕士学位论文][D].南京：东南大学,2005.
[10]江佩颖,李蔡彦等.以脸部特征为基础之夸张肖像画产生系统：[硕士学位论文][D].台湾：国立台湾海洋大学,2007.
[11]陈洪,郑南宁,梁林等.基于样本学习的肖像画自动生成算法系统.计算机学报[J].
[12]万里红,董兰芳,沈育平,陈意云.人物肖像画及其卡通动画仿真系统的建立.计算机仿真[J],2008,44(15)：71～72.
[13]张翼,陈纯.基于视觉特征提取的表情人脸动画生成：[硕士学位论文][D].杭州：浙江大学,2008.
[14]刘振安.基于SVG的卡通人脸图形自动生成法.测控技术[J],2006,vol(5)：101-107.
[15]林维训,潘纲,吴朝晖,潘云鹤.脸部特征定位方法综述.中国图象图形学报[J],2006.5,vol(5)：101-107.
[16]吴证.基于样本学习的人像线条画生成系统：[硕士学位论文][D].上海：上海交通大 学,2007.
[17]范玉华.基于ASM的人脸面部关键特征点定位算法研究：[硕士学位论文][D].郑州：河南科技大学,2008.
[18]曾阳.人脸特征提取方法研究：[硕士学位论文][D].长沙：中南大学,2008.
[19]吕宝华.人脸特征定位方法研究：[硕士学位论文][D].西安：西北工业大学大学,2007.
[20]吴证.人脸特征点定位及其应用：[硕士学位论文][D].上海：上海交通大学,2007.
[21]Cootes T F. Locating Faces Using Statistical Feature Detectors[C]. In:Proceedings of 2nd International Conference on Automatic Face and Gesture Recognition, Killington, 1996,204 —209.
[22]Thaddeus Beier. Feature-based Image Metamorphosis. Computer Graphics[J].1992, Vol.75:3-24.
[23]HER TZMANN A. Painterly Rendering With Curved Brush Strokes of Multiple Sizes[C]. In:Proceedings of the 25th Annual Conference on Computer Graphics and Interactive Techniques,1998.
[24]Ergun Akleman. Making Extreme Caricatures with a New Interactive 2D Deformation Techique With Simplicial. Complexes. Computer Graphics[J].1992, Vol.75:3-24
[25]Son Lam Phung, Douglas Chai, Bouzerdoum. Skin Color Based Faee Detection[C].In:Proeeeding of Intelligent Information Systems Conference.Australian and New Zealand,2001,171—176.
[26]HeiseleB, HoP, Poggio T. Face recognition with support veetor machines:global versus component-based approach[C].In:Proeeeding of International Conference on Computer Vision. Vancouver,Canada,July,2001,2,688—694.
[27]Yang G, Huang T S. Human face detection in a complex background[J]. Pattern Recognition,1994,27 (1):53—63.
[28]Feng G C, Yuen P C. Variance Projection function and its application to eye detection for human face recognition[J]. Pattern Recognition Letters,1998,19 (9):899—906.
[29]Yuille A L, Cohen D S, Halinan P W. Feature extraction from faces using deformable templates[C].In:Proeeedings of IEEE Computer Soc. Conf. on Computer Vision and Pattern Recognition. Washington,DC,USA,1989,104—109.
[30]Cootes T F, Taylor C J, CooPer D H,et al. Active shape models-their training and application[J].Computer Vision and Image Understanding,1995,61 (1):38—59.
[31]Hsu R L, Mottaleb M A, Jain A K. Face detection in color Images[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence,2002,24 (5):696—706.
[32]Waite J, Vincent J. A Probabilistic framework for neural network facial feature location[J].British Telecom Technology Journal,1992,10 (3):20—29.
[33]Turk M, Pentland A. Eigenfaces for recognition[J]. Journal of cognitive Neuroscience,1991,3 (1):71—86.
[34]Pan G, Lin W X, Wu Z H,et al. An eye detcetion system based on SVM filter[C].In:Proeeedings of Electronic Imaging and Multimedia Technology. Shanghai,China,2002,326 — 331.
[35]PJ.PhilliPs,H.Moon,etc. "The FERET Evaluation Methodology of Face-Reeongition Algorithms',IEEE Transactions on PAMI,VOL.22,No.10,pp 1090-1104,2000.
[36]BeierT, Neely S. Feature-based image metamorphosis[A].Poreeedings of ACM SIGGRAPH'92[C].Chicago:ACM Press,1992.35—42.
[37]RuPercht D,Mller H. Image warping with scattered data interpolation[J].IEEE Computer Graph Appl,1995,15:37—43.
[38]Hassnaien A E, Nakajima M. Image morphing of the facial image transformation based on Navier elastic body splines. Geneva,Switzerland,IEEE Computer Society Press,Los Alamitos,CA,pp 119—125.
[39]LeeS,ChwaK-Y,Hahn J, Shin SY (1996).Image morphing using deformation techniques.J Visualization Comput Anim 7:3-23.
[40]Gao P, Sederberg T W. A work minimization approach to imagemorphing[J].Visual ComPuter,1998.
[41]Michael Kass, Andrew Wiktin, and Demetri Terzopoulos. Snkaes:Active contour models. Intermational Joural of Computer Vision[J],pages 321—331.