CASIA OpenIR  > 毕业生  > 硕士学位论文
Alternative Title3D Face Modeling and Expression Synthesis from a Single Image
Thesis Advisor刘昌平 ; 黄磊
Degree Grantor中国科学院大学
Place of Conferral中国科学院自动化研究所
Degree Discipline计算机应用技术
Keyword人脸建模 表情合成 主动形状模型 散点插值 运动模型 面部划分 Face Modeling Expression Synthesis Asm Scattered Data Interpolation Motion Model Facial Region Segmentation
Abstract具有真实感的三维人脸合成是计算机视觉与计算机图形学领域具有挑战性的研究课题之一,并因为其在工业界的多个领域中的良好应用前景而被广泛研究。三维人脸合成技术有多个分支,其中最有趣且最具有挑战性的研究课题是基于单幅图像的人脸合成问题。 在对人脸合成的研究背景、主要难点和技术路线等进行详细的研究的基础之上,本文回顾了基于单幅图像的三维人脸建模与表情合成的经典问题与处理方法,并且实现了一个新型的三维人脸表情合成系统。 本文实现了由单幅图像进行人脸建模的系统。使用单幅图像作为输入,利用主动形状模型进行人脸的特征定位与形状捕获,在形状约束下使用径向基函数对一个标准三维人脸模型进行散点插值以完成形状建模,并使用形状约束对模型进行纹理映射。 文章提出了一种新的数据驱动的表情合成方法。该方法基于人脸建模系统和一个有限三维表情模型集,是本文的主要贡献。该系统使用表情运动模型框架,将表情合成问题转化为表情的三维运动特征建模问题。为了使得运动模型更加适应输入图像中的人脸形状,我们提出了一种局部变换方法对表情的运动进行变换,在此基础上构造了线性表情运动模型,使得系统能够实时合成出具有真实感的人脸表情。 此外,本文提出了一种新型的基于聚类的面部区域划分方法。不同于以往的人工面部划分等方法,我们利用人脸在各个表情中的运动特征,将面部区域进行自动的划分。区域划分模块能够丰富系统输出,使得系统能够合成出输入表情集中所不含有的多种人脸表情。
Other AbstractRealistic 3D face modeling and expression synthesis are challenging topics in the research field of both computer vision and computer graphics, which have been extensively researched for their application prospect in the industry. Among all those types of face synthesis researches, single image-based face synthesis is the most interesting as well as the most challenging one. On the basis of the research on the background of face synthesis, this paper focused on the face modeling and facial expression synthesis from a single image. Meanwhile, a new approach for 3D facial expression synthesis was proposed in this paper. A single-image-based face modeling procedure was implemented in this paper. On the basis of the single image input, we employed the ASM to capture the feature of the face. The result of ASM and a reference model were adopted to form a shape modeling process which was implemented by scattered data interpolation using RBF. The face modeling was finalized with texture mapping. The main contribution of this paper is a novel approach for data-driven facial expression synthesis, which was build on the basis of the face modeling and a finite 3D facial expression set. The approach is constructed under an motion model framework. In this framework a facial expression was synthesized with the corresponding motion features and a neutral face. A local transform procedure was proposed to fit the motion model to the shape of the input face. After that, we constructed the linear expression motion models which enable the system to synthesize the realistic facial expressions in real time. In addition, a novel clustering-based approach for facial region segmentation was proposed. Unlike other manual ones, this approach run automatically using the motion feature of the face among the expressions. With the segmentation, the system can generate many expressions which are not existed in the expression set.
Other Identifier201028014629075
Document Type学位论文
Recommended Citation
GB/T 7714
舒之昕. 一种基于单幅图像的三维人脸建模与表情合成方法[D]. 中国科学院自动化研究所. 中国科学院大学,2013.
Files in This Item:
File Name/Size DocType Version Access License
CASIA_20102801462907(2317KB) 暂不开放CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[舒之昕]'s Articles
Baidu academic
Similar articles in Baidu academic
[舒之昕]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[舒之昕]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.