基于统计模型与Gabor小波的人脸对齐

Face alignment based on statistical models and Gabor wavelet

将基于Gabor小波的人脸特征点跟踪算法与基于统计模型的主动外观模型AAM人脸特征点定位方法结合起来,实现视频中人脸的自动对齐。先利用Gabor小波进行特征点跟踪,其结果作为AAM的初始形状。利用AAM的全局形状和纹理信息作为约束,对Gabor小波的局部跟踪错误进行校正。实验表明,该方法是有效的。

A face alignment scheme was proposed by integrating the Gabor-wavelet facial feature points tracking algorithm with the facial feature points localization method based on statistical mode AAM(Active Appearance Model). First the facial feature points were tracked with Gabor wavelet, then the result was used as the initial shape of the AAM. The local tracking errors of Gabor wavelet were corrected with the global constrained shape and texture information of AAM. Experiments prove that the method works well.