面向模型基人脸视频编解码领域,提出了一种基于MPEG-4的三维人脸表情动画算法.首先对编码端发送视频的首帧图像,利用Adaboost+Camshift+AAM(active appearance model算法检测人脸和定位特征点,接着特定化一个简洁人脸通用网格模型得到FDP(facial definition parameter);对于得到的FDP,解码端先用其特定化一个精细人脸通用网格模型,然后基于肌肉模型和参数模型相结合的方式来生成人脸表情动画,同时对人脸功能区进行划分.实验表明,该算法在FAP(facial animation parameter)流的驱动下可以生成真实感较强的三维人脸表情动画.%In view of the model based coding/decoding area, a 3D facial expression animation system based on MPEG-4 was proposed. The coder obtained FDPs (facial definition parameter) through face adaptation of a simple universal triangular model with Adaboost + Camshift + AAM algorithm for face detection and feature localization in the first frame. Then the decoder adapted a complex universal triangular model using these FDPs, Finally the algorithm produced facial animation combining the parameterized model and muscle model. A facial action area split scheme was also proposed. Experiment results confirm that this system can produce realistic facial expression animation with FAP (facial animation parameter) flow.
展开▼