全自动实时人脸图像变形技术研究与实现

Research and Implementation of Automatic and Real-time Face Morphing Approach

提出了一种新的人脸图像变形方法——自适应网格方法(AMM)。该方法可以全自动产生实时的人脸变形效果。ASM(主动形状模型)技术的使用消除了其他人脸变形算法中所必须的手工标定人脸关键点工作;提出一种将人脸矩形区域自动划分成数个特征矩形的方法,使得复杂的图像变形计算被化简为简单高效的图像插值计算。该工作有助于将人脸变形技术运用于桌面实时应用中。

This paper presents a .novel approach for face image inorphing Adaptive Mesh Method (AMM) , which provides automatic and real-time metamorphosis between pairs of face images. Applying ASM (Active Shape Model) technology, human intervention for feature specification is eliminated by the automatic facial landmark detection. In order to acquire real-time performance, a new method, which can automatically subdivide face into rectangular meshes is presented, thus simplifies the complex image-warping algorithm to simple and effective interpolating algorithm. Our work makes it possible to use face-morphing technology in real-time application on desktop.