基于贝赛尔曲面的三维人脸特征点定位方法

三维人脸识别技术中,人脸特征点的精确定位和匹配对后期的人脸识别起了极为关键的作用,为此提出了一种新的三维人脸特征点定位方法,利用矩阵特征分解求解每个特征点的基准平面,进而可用贝塞尔曲面来表达特征点附近的局部三维表面结构,对不同复杂结构的表面可用不同的精度描述,采用局部搜索和全局形状模型的约束相结合得到三维人脸特征点的精确定位.对三维人脸的特征点进行定位的实验结果证明,该方法比其他基于曲率分析的方法具有更精确的定位精度.

Three Dimensional Localization of Facial Feature Points Based on Bezier Surface

In 3D face recognition methods,precise localization of facial feature points plays an important role in later facial recognition.A novel localization method of 3D feature points is presented.A uniform plane is first calculated by solving eigen-decomposition problem,then Bezier surface is used to represent local structure around different feature points.In addition,different interpolation accuracies are adopted for different complexities of feature points.Local search and global constrain are combined to get the accurate localization.Experimental results on 3D faces indicate that the present method achieves higher accuracy than traditional curvature-based methods.