摘要
基于表情活动单元的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。本文提出的算法使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。本算法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MissAmerica(CIF ,10Hz)上实验 ,证明了其正确性和有效性。
The automatic adaptation of a face model using action units is necessary for the model based coding of videophone sequences at very low bit rates.In the proposed algorithm,eye and mouth features are firstly estimated using deformable templates.Then,the face model Candide is adapted to these estimated features.The algorithm differs from previous works in the following aspects:there is no restriction on the rotation for the global adaptation,and initial values of action units are determined due to the mimic adaptation.The algorithm has been applied to the test sequences Akiyo and Miss America (CIF,10 Hz) with high accuracy and efficiency.
出处
《南京邮电学院学报》
2000年第2期6-10,21,共6页
Journal of Nanjing University of Posts and Telecommunications(Natural Science)
基金
信息产业部中青年项目
关键词
图像编码
表情活动单元
人脸模型
自动适配
Image coding
Action units
Face model
Model based coding
Semantic coding
