摘要
提出了一种基于标注图像的MPEG-4人脸运动参数(FAP)获取方法。首先对人脸进行手工标点,然后使用摄像机拍摄人脸运动序列并录制语音数据,通过图像分割与标记点定位等操作提取标记点位置信息,再根据人脸特征点的几何分布确认标记点对应的三维FDP参数,最后经过计算得到人脸运动参数。将获取的结果数据应用到基于MPEG-4的语音驱动的三维人脸合成与动画系统中,取得了很好的效果。
In this paper we put forward a method through which we can obtain the MPEG-4 FAP (Facial Animation Parameter). Firstly, we put some marks on the human face according to the MPEG-4 FDP (Face definition parameter), then we record the video sequence of the face movement and voice data. By segmenting and locating the marked points in the images, we extract location of the feature points marked on the human face and then confirm them according to the pre-known knowledge of feature points location of human faces. Finally we get the FAP (Facial Animation parameter). We obtained very good result when the result data was applied to the synthesis and animation of human faces.
出处
《系统仿真学报》
CAS
CSCD
2001年第S2期497-501,共5页
Journal of System Simulation
基金
国家自然科学基金重点项目(69789301)
国家863计划项目(863-306-ZD03-01-2)
中科院百人计划的资助。
