基于ASM和ERT特征点定位算法的人脸比例特征信息提取

Extraction of Human Facial Proportions Based on ASM and ERT Feature Point Location Algorithms

提出一种基于ASM和ERT特征点定位算法的人脸比例特征信息提取方法。选用图像几何特征和人眼的位置特征,将人脸图像转化为用向量表示的数学符号,这些特征向量基本上可以反映出整张人脸图像的有效信息,其中人脸的比例特征和人眼的位置特征用到了现阶段的两个主流的特征点定位算法,即ASM和ERT,然后利用人脸的五官比例去量化得到8维的特征信息。文中还从时间性能和准确率方面对两种算法进行对比,结果表明算法精确度上ERT略优与ASM,算法时间上PC端测试ERT明显优于ASM。

This paper proposes a method of extracting facial features based on ASM and ERT feature point localization algorithms. Using the geometrical features of the face and the position features of the eyes, the face image is transformed into a mathematical symbol represented by vectors. These eigenvectors can basically reflect the effective information of the whole human face, in which the proportionality of the face and the eyesisthe two main feature point localization algorithms, namely ASM and ERT, and then we use the facial features to quantify the 8-dimensional feature information. In this paper, we also compare the two algorithms from the time and accuracy. The results show that the ERT is better than ASM in the algorithm accuracy.