文档介绍:
一种基于 DPM 的人脸特征点定位方法
范文萍,苏菲**
(北京邮电大学信息与通信工程学院,北京 100876)
5
10
15
20
25
摘要:人脸特征点定位在人脸识别的相关领域应用十分广泛。例如人脸识别中常借助特征点
来进行人脸对齐,人脸 3D 建模常利用特征点来进行图像重建。本文给出了一种基于 DPM 模
型的人脸特征点定位方法,该方法首先提取人脸图像中特征点邻域图像的纹理特征及特征点
间的形状特征,利用结构化输出分类器进行参数训练,最后利用动态规划的方法进行人脸特
征点定位。实验结果表明,该算法可以快速准确地定位人脸特征点。
关键词:人脸识别;人脸特征点定位;DPM;LBP
中图分类号:
A face landmarks detection method based on Deformable
Part Model
FAN Wenping, SU Fei
(School of Information munication Engineering, Beijing University of Posts and
munications, Beijing 100876)
Abstract: Face landmarks detection has been widely used in face recognition related fields. For
example, it can be utilized to normalize face image in face recognition system, and reconstruct
image in 3D face modeling. In this paper, a novel face landmarks detection method based on
deformable part model (DPM) is proposed. In this algorithm, landmarks detector is defined as a
structured output classifier, whose input are landmarks' local appearance feature and shape feature
of face, and the output is the positions of face landmarks. This classifier is trained to get the
optimal parameter vector to detect face landmarks by implementing dynamic programming. The
experimental results on LFW database show that this method performs well both in execution
efficiency and accuracy.
Key words: Face recognition; Face landmarks detection; Deformable Part Model; Loc