基于特征发现的卡通人脸肖像生成  

Caricature Generation Based on Facial Feature Analysis

作　　者：周仁琴[1] 周经野[1] 陈益强[2] 刘军发[2] 

机构地区：[1]湘潭大学信息工程学院计算机科学与技术系 [2]中国科学院计算技术研究所先进人机通信技术联合实验室

出　　处：《计算机辅助设计与图形学学报》2006年第9期1362-1366,共5页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60303018;60473043);北京市科技新星计划项目(2005B54)

摘　　要：通过对成年男女各100幅真实照片进行特征提取和特征统计,获得平均人脸特征分布信息,对新输入的人脸照片进行特征比对,发现其相对突出的特征,采用主动形状模型特征提取和特征线对相结合的方法,对突出的特征实现自动变形,生成人物的卡通肖像.实验结果表明,该方法具有人脸数据量大、特征提取和发现的自动化、变形效果好等优点.This paper presents a system for automatic generation of human caricature. By the scheme proposed in the system, we first extract facial feature parameters from the real facial photos of 100 males and 100 females respectively, and we calculate the average facial features through a statistic procedure. There are in total 19 features parameters with their distributions calculated. Then when a new facial image is inputted, its features are compared with the average face to discover the personal features of the individual, and by exaggerating the personal features, we realize the caricature through image warping. In the warping process, the improved active shape model （ASM） and feature based warping method are adopted, and as a result, we can extract the features and warp the face automatically. Experiments show that our method performs well.

关 键 词：漫画 肖像生成 图像处理 图像变形 人脸动画 卡通动画 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]