基于光流和Gabor小波的人脸特征点跟踪

摘要

This paper studies a facial feature points tracking method. This approach solves the problems that the traditional optical flow, Snake, and deformable template methods can not effectively track the variation in texture characteristics of the feature points. This paper presents a method to divide the feature points into different groups, and each group is tracked with improved optical flow method and the elastic graph matching method. Fourteen points around the eyes and eyebrows are tarcked by optical glow. Because the expression of eight feature points near the mouth is rich, gabor wavelet is used to track these points. Experimental results show that this method has good effect and is conducive to the promotion.%研究优化人脸特征提取问题,针对长期以来在不贴标记点的情况下用传统的光流、Snake、可变模板等方法对纹理特征变化大的特征点不能有效跟踪,并且解决单独采用Gabor小波系统开销大等问题,为了在人脸图像中提取准确信息,提出了人脸特征点的跟踪方法,分组采用改进的光流法和弹性图匹配的方法进行特征点跟踪.对眼睛、眉毛、上下眼皮等14个表情变化不大的特征点使用光流法进行跟踪,最后对变化大的嘴部8个特征点运用Gabor小波的弹性图匹配方法进行仿真,实验结果表明获得了很好的跟踪效果,且利于推广使用.