人脸动画中基于样条插值的器官拼接算法实现

The Implementation of Face Organ Mosaic Algorithm Based on Spline Interpolation

人脸动画广泛地应用于游戏行业、远程会议、代理和化身等许多其它领域,近年吸引了很多学者的研究,其中口腔/眼睛等器官的动画一直是一个较大的难点。本文提出了一种将口腔/眼睛的器官样本图片融合到人脸图像中并根据单张中性人脸图片生成人脸动画的方法。该方法根据特征点生成样条,在极坐标上对样条插值来实现空间映射,然后采用后向映射和插值进行图像重采样得到融合图像。实验结果表明,该方法产生的融合图片较为自然,能实现口腔/眼球等器官的运动,能满足人脸动画生成的实时性要求。

Facial animation which attracts many researchers recent years is widely used in the fields of game industry, teleconferencing, agent and avatar, in which the animation of mouth/eye organs is always a tough issue. This paper presents an image mosaic algorithm to fuse the mouth/eye organ sample images into face image and generate the facial animation based on one neutral facial image. In this method, spline generated by facial feature points is used for image spatial mapping on polar coordinates. Then resampling the image by using inverse mapping and interpolation to get the fused image. Experiments show that the fused image generated by this approach is relatively natural. It can implement the mouth/eye animation and satisfy the real-time requirement of facial animation.