多表情源的人脸表情合成技术

提出了基于一个局部约束偏移和相对偏移技术的图像变形方法，用于人脸表情合成．该方法可以自动克服目前表情合成方法的非特征区域的形变，并与脸型外观相匹配．在此基础上，提出了基于多表情源的人脸表情合成技术．从人脸图像划分出三个主要表现表情的区域，并对每个区域根据多个表情源的加权平均的相对偏移量进行图像局部变形，最后利用加权平均的表情比图像加强表情．实验结果表明，用该方法得到的人脸表情自然逼真，更为重要的是，通过调节权向量的取值，可以取得不同模式的表情，丰富了表情表现的内涵．

Facial Expression Synthesis from Multiple Facial Expressions

In this paper, we propose two approaches of image warping, applied to facial expression synthesis, respectively by locally constrained offset forming and relative offsetting. These modifications can reduce unnecessary offset outside the character part and match different face figures. Based upon the methods, we present a facial expression synthesizing method from multiple facial expressions. Experimental results show that facial expression images produced by the method are quite natural and lively. Most importantly, this method can produce different facial expressions by adjusting its weigh vectors.