摘要
人脸模型自动适配是极低码率可视电话模型基编码所必需的算法。该算法用变形模板估计眼睛和嘴巴的特征 ,利用这些特征用 Candide模型对人脸进行适配 ,它对人脸模型的全局适配没有旋转限制 ,而且表情活动单元的初值决定于模仿适配。算法在视频序列 Akiyo和 Miss America(CIF,1 0 Hz)上实验 ,证明了其正确性和有效性。
Automatic adaptation of face model is necessary to model based coding of videophone sequences at very low bit rates A new algorithm for automatic adaptation is proposed First,eye and mouth features are estimated by means of the algorithm using deformable templates Then, these features are adapted by Candide model The algorithm is different from previous works in the following aspects:①there is no restriction on the rotation of the mode that is global adaption, and ②initial values of action units are dependent on the mimic adaptation The algorithm is proved more accurate and effective by application of the test sequences Akiyo and Miss America (CIF,10Hz)
出处
《江汉石油学院学报》
CSCD
北大核心
2000年第2期67-70,共4页
Journal of Jianghan Petroleum Institute
关键词
图像识别
人脸模型
可视电话
自动适配方法
算法
pattern recognition
action unit
face model
model based coding
semantic coding