[1]汪丰,祁长红,刘成,等.基于正交图像的头部三维模型构建[J].东南大学学报(自然科学版),2015,45(1):36-40.[doi:10.3969/j.issn.1001-0505.2015.01.007]
 Wang Feng,Qi Changhong,Liu Cheng,et al.Reconstruction of 3D head model based on orthogonal images[J].Journal of Southeast University (Natural Science Edition),2015,45(1):36-40.[doi:10.3969/j.issn.1001-0505.2015.01.007]
点击复制

基于正交图像的头部三维模型构建()
分享到:

《东南大学学报(自然科学版)》[ISSN:1001-0505/CN:32-1178/N]

卷:
45
期数:
2015年第1期
页码:
36-40
栏目:
计算机科学与工程
出版日期:
2015-01-20

文章信息/Info

Title:
Reconstruction of 3D head model based on orthogonal images
作者:
汪丰祁长红刘成姜伟倪舟邹亚
东南大学生物科学与医学工程学院, 南京 210096
Author(s):
Wang Feng Qi Changhong Liu Cheng Jiang Wei Ni Zhou Zou Ya
School of Biological Sciences and Medical Engineering, Southeast University, Nanjing 210096, China
关键词:
头部模型重建 正交图像 Candide-3模型 纹理映射 主动形状模型
Keywords:
head model reconstruction orthogonal image Candide-3 model texture mapping active shape model
分类号:
TP391
DOI:
10.3969/j.issn.1001-0505.2015.01.007
摘要:
为快速构建人体头部的三维模型,提出了一种使用头部正面和侧面两张正交图像进行重建的方法.首先,利用Candide-3模型生成标准头部模型;然后,利用改进的主动形状模型(ASM)自动获取正面图像中68个特征点和侧面图像中35个特征点,并将获取的特征点与标准头部模型中的顶点进行匹配,利用匹配特征点对标准模型进行全局变形和局部变形,从而获得个性化模型;最后,对侧面头像进行仿射变换,使之与正面图像配准,并通过柱面投影和加权融合获得人脸全景纹理图,将其映射到个性化模型上以获得逼真的头部模型.实验结果表明:与传统ASM算法相比,利用改进ASM算法获得的特征点定位准确度提高了13.1%;所提方法仅需约1.2 s便可重建出具有真实感的头部三维模型,并能提供较大范围的可视角度和较丰富的深度信息.
Abstract:
To quickly build a three dimensional head model, a reconstruction method using a frontal orthogonal image and a profile orthogonal image is proposed. First, the standard head model is generated by the Candide-3 model. Then, 68 feature points and 35 feature points are automatically acquired from the frontal and profile images respectively by using the improved active shape model(ASM). The obtained feature points are matched with the vertexes of the standard head model. The specific head model is generated by deforming the standard head model globally and locally with the matched feature points. Finally, affine transformation is carried out on the profile image for image registration with the frontal image. The registered profile image is obtained by cylindrical projection and weighted fusion. The texture map is mapped on the specific head model to get a realistic head model. The experimental results show that the positioning accuracy of the feature points by using the improved ASM is increased by 13.1% compared with that by using the classical ASM. By using the proposed method, building the realistic head model with a wide range of viewing angle and rich depth information only needs about 1.2 s.

参考文献/References:

[1] Pighin F, Hecker J, Lischinski D, et al. Synthesizing realistic facial expression from photographs[C]//Proceedings Computer Graphics and Interactive Techniques. Los Angeles, CA, USA, 1998:75-84.
[2] Fooprateepsiri R, Kurutach W. A general framework for face reconstruction using single still image based on 2D-to-3D transformation kernel[J]. Forensic Science International,2014, 236:117-126.
[3] Heo J, Savvides M. Rapid 3D face modeling using a frontal face and a profile face for accurate 2D pose synthesis[C]//Proceedings of Automatic Face and Gesture Recongnition and Workshops. Santa Barbara, CA, USA,2011: 632-638.
[4] Ahlberg J. Candide-3—an updated parameterized face[R]. Linkoping, Sweden: Department of Electrical Engineering of Linkoping University, 2001.
[5] Berkay Y M, Erdogan H, Unel M. Facial feature extraction using a probabilistic approach[J]. Signal Processing: Image Communication, 2012, 27(6):678-693.
[6] Yurtkan K, Demirel H. Feature selection for improved 3D facial expression recognition [J]. Pattern Recognition Letters, 2014, 38:26-33.
[7] Arámbula Cosío F, Flores J A, Castañeda M A. Use of simplex search in active shape models for improved boundary segmentation[J]. Pattern Recognition Letters, 2010, 31(9):806-817.
[8] Solina F, Peer P, Batagelj B, et al. Color-based face detection in the “15 seconds of fame” art installation[C]//Proceedings of Mirage. Rocquencourt, France, 2003:38-47.
[9] Louis W, Plataniotis K N. Frontal face detection for surveillance purposes using dual local binary patterns features [C]//IEEE International Conference on Image Processing. Hong Kong, China, 2010:3809-3912.
[10] Zhou Z, Li S, Wang B. Multi-scale weighted gradient-based fusion for multi-focus images[J].Information Fusion,2014, 20:60-72.
[11] Romdhani S,Vetter T. Efficient, robust and accurate fitting of a 3D morphable model[C]//Proceedings of the Ninth IEEE International Conference on Computer Vision. Nice, France, 2003:59-66.
[12] Levine M, Yu Y. State-of-the-art of 3D facial reconstruction methods for face recognition based on a single 2D training image per person[J].Pattern Recognition Letters, 2009, 30(10): 908-913.

备注/Memo

备注/Memo:
收稿日期: 2014-07-08.
作者简介: 汪丰(1969—),男,博士,副教授,feng.wang@seu.edu.cn.
引用本文: 汪丰,祁长红,刘成,等.基于正交图像的头部三维模型构建[J].东南大学学报:自然科学版,2015,45(1):36-40. [doi:10.3969/j.issn.1001-0505.2015.01.007]
更新日期/Last Update: 2015-01-20