J4 ›› 2012, Vol. 30 ›› Issue (6): 605-609.

• 论文 • 上一篇    下一篇

基于形变模型的三维人脸快速重建改进算法

胡阳明, 周大可, 鹿乐, 杨欣   

  1. 南京航空航天大学 自动化学院, 南京 210016
  • 收稿日期:2012-08-14 出版日期:2012-11-23 发布日期:2013-06-05
  • 作者简介:胡阳明(1988—), 男, 江苏盐城人, 南京航空航天大学硕士研究生, 主要从事三维人脸重建与识别、 目标跟踪与识别研究, (Tel)86-15895972742(E-mail)huyangmingfly@126.com|周大可(1974—), 男, 江苏涟水人, 南京航空航天大学副教授, 硕士生导师, 主要从事生物特征识别、 计算机视觉、 图像处理研究,(Tel)86-25-84892305-5112(E-mail)dkzhou@nuaa.edu.cn。
  • 基金资助:

    国家自然科学基金资助项目(61172135); 南京航空航天大学基本科研业务费专项科研基金资助项目(NS2010089)

Improved 3D Fast Face Reconstruction Algorithm Based on Morphable Model

HU Yang-ming, ZHOU Da-ke, LU Le, YANG Xin   

  1. College of Automation Engineering, Nanjing University of Aeronautics and Astronautics, Nanjing 210016, China
  • Received:2012-08-14 Online:2012-11-23 Published:2013-06-05

摘要:

针对传统三维人脸重建算法效率低且难以满足实际应用的缺陷, 提出一种三维人脸重建改进算法。该算法基于ASM(Active Shape Model) 增强算法, 自动地对特定二维人脸特征区域进行准确定位, 并实现三维人脸数据库的归一化; 利用稀疏形变模型对特定正面二维人脸进行快速三维重建; 采用明暗纹理恢复算法对重建后的三维模型人脸特征区域的每个顶点法线进行约束, 并将其应用于人脸识别中。实验结果表明, 该方法可实现对特定正面二维人脸快速三维重建, 并取得较好的三维重建精度与识别率, 与经典ASM算法相比, 精度提高12.3%, 迭代次数减少6次。

关键词: 人脸三维重建, 形变模型, 明暗纹理恢复, 人脸识别, 主动形状模型

Abstract:

The traditional algorithm of 3D face reconstruction is inefficient and difficult to meet the requirements of practical application. To address this problem, a 3D face fast reconstruction algorithm is proposed. The algorithm based on improved ASM (Active Shape Model) is used to automatically and accurately positioning the feature points, and the improved ASMbased algorithm is used to normalized 3D face database. The sparse 3D morphable model is established to fast reconstruction 3D face from the 2D frontal face; Finally, shading information of 2D image is fully used as a prior knowledge to constrain each vertex's normal located in feature region of 2D face. Experimental results certify that the proposed method obtains faster speed to reconstruction 3D face from the 2D frontal face. keeping a quite high accuracy results and having a high recognition rate.

Key words: 3D face reconstruction, morphable model, shape from shading(SFS), face recognition, active shape model(ASM)

中图分类号: 

  • TP391