基于优化的SIFT特征描述子的人脸特征点定位

Precise Facial Landmark Localization Based on Optimal SIFT-based Feature Descriptors

针对传统人脸特征点定位算法复杂度高,精确度低和适应性差的特点,提出了1种基于脸部特征点特有纹理特征的检测方法进行精确快速的人脸特征点定位.首先根据脸部特征点的纹理特征,利用Powell算法学习得到基于SIFT(Scale Invariant Feature Transform)的特征算子最优化的参数.然后提取脸部特征点在最优化的参数下的SIFT特征算子并用于训练基于支持向量机回归的检测器.最后利用Bio ID人脸数据库进行测试.实验结果表明,该方法结构简单,具有较高的精确度,对于表情和光线变化具有很好的鲁棒性.

Considering the high computational complexity, low accuracy and poor adaptability of tradi- tional facial landmark localization methods, a texture-based method is proposed to do the precise landmark localization in the wild dataset. Firstly, it obtained the parameters that optimize the Scale Invariant Feature Transform （SIFT）-based landmark-specific feature descriptors in accordance with their special textural char- acteristics, by adopting the Powell algorithm. Secondly, it trained support vector machine （SVM） regressors with the landmarks＇ SIFT-based feature descriptors calculated based on those learned parameters. At last, it tested the method on the widely used benchmarks which is BioID database. The results manifest that this method is not only simple and more accurate, but also robust to the expression and illumination.