实现人脸图像变形算法的研究

Study on Realization of Person's Face Image Deformation

利用主动形状模型(ASM,Active Shape Model)算法进行人脸的68个关键特征点提取,得到了输入图像人脸68个特征点分布规律和具体的位置。研究和改进了线性的最小移动二乘法(MLS),重点研究了变形函数中转换矩阵为刚性变换的方法,在进行可控的控制点变形方法过程中,MLS在进行刚性变形中能保持图形的总面积不变,在结合人脸表情和特征点位置变化规则,对有弹性物体进行图像的变形,实验结果表明,产生的图像变形效果更加具有真实感。

The active-shape-model ASM algorithm is employed to extract the 68 key characteristics of person＇s face,and thus to obtain the distribution regulations and positions of these 68 key characteristic points.The linear moving least squares（MLS） is studied,the point inquiries into the concerning control point can control the transform method,and MLS could keep the total area of the sketch constant in the rigid transform.And for the transform of flexible object,the produced transform result is of much more sense-of-reality.