在人脸序列的图象编码中,由于模型基编码方法可以获得高的主观图象质量和低的码率,因而受到广泛重视.但是,其运动参数的可靠估计还是一个难点.为此,该文分析了头部运动的特点,并把它分为头部刚体运动、脸部表情的简单运动和脸部表情复杂运动3种形式.其中,提取头部刚体运动参数利用了基于特征点对的运动参数估计算法,并提出了一个线性的实现方法;文中还指出提高运动参数估计的精度在于选择合适的特征点和建立一个和特定人脸相一致的三维线框模型;另外,还为脸部表情的简单运动建立了形变矩阵;最后给出了用面积误差函数评价的运

Model-based facial image coding has received extensive attention due to its high subjective image quality and low bit-rates. But the estimation of object motion parameters is still a difficult problem. In this paper, head motion characteristics is analyzed, and it can be classified into three motion types, head rigid motion, simple motion of facial expression and complex motion of facial expression. A motion parameter estimation algorithm based on corresponding feature points is utilized to extract head rigid motion. A linear implementation algorithm is suggested. The key works to improve the accuracy of motion parameter estimation are the selecting of some proper feature points and the building of an exactly 3D wire-frame corresponding to specified face. A deformable matrix is constructed to extract simple motion parameter of facial expression. At last, the error of motion parameter is evaluated by the area-error function.