三维卡通人脸的样本集构建及其个性化生成的研究

发布时间：2018-02-14 20:41

本文关键词： 计算机图形学三维卡通人脸流形学习主成分分析　出处：《湘潭大学》2009年硕士论文　论文类型：学位论文

【摘要】： 在信息技术迅猛发展的今天,数字娱乐已融入到人们生活的各个角落。近年来,卡通产品受到了社会各年龄阶层人群的青睐。随着三维虚拟环境技术的发展和普遍应用,三维卡通形象在动漫影视、在线游戏、虚拟社区、辅助教学等领域呈现出越来越广泛的应用。目前,传统的手工制作方法和基于规则生成的方法都无法解决实际应用中实时性、艺术背景相关性、制作成本过大等问题。本文采用流形学习、主成分分析等方法对三维卡通人脸生成的问题展开研究,取得了如下的研究成果: 首先本文建立了一个由二维卡通人脸图片、三维卡通人脸模型组成的卡通样本库,进一步地,本文对三维人脸部件给出了定义,从而相应地也构建出了三维卡通人脸部件样本库。其次,本文应用局部线性嵌入的降维方法来获得样本库中二维卡通和三维卡通人脸的低维嵌入,然后再应用半监督流形学习的方法挖掘出二维卡通数据集与三维卡通人脸数据集之间的映射关系,将此映射关系应用于二维卡通来获得与之对应的三维卡通人脸的低维嵌入,再通过局部线性嵌入的升维方法重构出其相应的三维卡通人脸,从而扩充三维卡通人脸样本库。最后本文基于扩充的三维卡通人脸样本库,采用主成分分析的方法实现了个性化的三维卡通人脸生成。
[Abstract]:With the rapid development of information technology, digital entertainment has been integrated into every corner of people's life. In recent years, cartoon products have been favored by people of all ages. The 3D cartoon image has been used more and more widely in the fields of animation, film and television, online game, virtual community, assistant teaching and so on. At present, the traditional manual production method and the rule based method can not solve the real time problem in the practical application. In this paper, manifold learning and principal component analysis are used to study the problem of 3D cartoon face generation, and the following research results are obtained:. First of all, a cartoon sample library composed of 2D cartoon face images and 3D cartoon face models is established. Furthermore, the definition of 3D face components is given in this paper. Accordingly, a 3D cartoon face sample library is constructed. Secondly, the method of local linear embedding is used to obtain the low dimensional embedding of 2D cartoon and 3D cartoon face in the sample database. Then the mapping relationship between 2D cartoon data set and 3D cartoon face data set is mined by semi-supervised manifold learning method, and the mapping relation is applied to two-dimensional cartoon to obtain the low-dimensional embedding of the corresponding 3D cartoon face. Then the corresponding 3D cartoon face is reconstructed by the method of local linear embedding, and the 3D cartoon face sample database is expanded. Finally, based on the expanded 3D cartoon face sample database, the method of principal component analysis (PCA) is used to realize personalized 3D cartoon face generation.
【学位授予单位】：湘潭大学
【学位级别】：硕士
【学位授予年份】：2009
【分类号】：TP391.41

【引证文献】