利用人脸特征及其关系的漫画夸张与合成

在人脸漫画中,人脸特征(脸型以及五官)形状的夸张与特征之间关系的夸张是存在差异的,而目前由计算机生成的人脸漫画算法中并未考虑这一点,为此提出一种漫画夸张与合成方法.该方法由基于样本的特征的形状夸张、基于规则的特征之间的关系夸张以及两者之间的合成3部分组成.在人脸特征的形状夸张部分不需要大量的样本库,只通过少量的漫画样本便可以捕捉形状的夸张;在人脸特征之间的关系夸张部分,引入了直观的"T型规则"方式描述人脸特征在夸张时的相互关系;在人脸形状夸张与关系夸张两者的合成部分,采用"比例"的形式来描述人脸特征以及特征之间的关系,使得夸张后可以方便地进行两者的合成.最后采用调查问卷的方式评测该方法生成的漫画效果,结果表明其效果良好.

Caricature Synthesis and Exaggeration Based on Facial Features and their Relationship

The difference between shape exaggeration and relationship exaggeration of face features has been seldom considered in existing algorithms of caricature generation.A new method for caricature exaggeration and synthesis is proposed in the paper,covering shape exaggeration,relationship exaggeration and global synthesis.Rather than relying on a large training set of caricature face pairs,the shape exaggeration step is based only on several caricature examples to capture the shape exaggeration.T-Shape rule is i...