基于特征网格的面部图象变形算法

A Distort Algorithm for Face Image Based on Character Mesh

面部图象变形技术是人脸建模的关键技术之一 .为了模拟人脸器官 (如眼、嘴、鼻和眉等 )周围及人脸轮廓肌肉的运动 ,选择适合描述人脸及其器官轮廓的几何形状的若干特征点 ,并根据器官周围肌肉点运动机理 ,建立了面部图象变形的特征网络模型 .该模型构造了两类控制器官运动的合成肌肉 ,即控制垂直方向运动的轮廓肌和控制水平方向运动的线性肌 ,由此 ,面部肌肉点拉动肌肉即可作垂直和水平方向运动 .由于该特征网络模型可实现人脸器官精细的变形操作 ,并能依据器官的轮廓变化展开 ,符合人的认知过程 ,因此 ,与其他算法相比 ,其变形效果更加自然 。

Morphing technique of human face image is a key technique in the field of face modelization. Based on geometrical feature of facial contour, this paper selects appropriate Feature Points and constructs a component feature mesh model adapted to face image morphing through the muscle movement principle of face organs. In order to imitate the muscle movement of facial component, such as eye, nose, brow, mouth and face contour, feature mesh constitutes two kind of synthetic muscle. One is the contour muscle which manipulates the removal of vertical direction. Another is the linear muscle that controls the displacement of horizontal direction. Thus, the movement of muscle point can be divided into two directions, vertical and horizontal, which is controlled respectively by the contour muscle and linear muscle where the point locates. Feature mesh model ensures that facial component morphing is manipulated in virtue of the change of component contour, coinciding with psychological cognition. This makes it more adapted to human image morphing than other morphing algorithms. The result of experiment shows that the facial images which generated by this algorithm are accurate, and the expression effects are more natural.