计算机技术与发展2015，Vol.25Issue(11)：1-5,5.DOI:10.3969/j.issn.1673-629X.2015.11.001
基于CPR和CLM的多视角人脸特征点定位方法
A Multi-view Facial Landmark Localization Method Based on CPR and CLM
摘要
Abstract
Facial landmark localization plays an important role in many face-related applications such as face recognition and 3D face re-construction. Although existing methods already achieve promising results on frontal and near-frontal face images,their performance on face images with large pose angles is still far from being satisfactory. A multi-view facial landmark localization method is proposed in this paper. It divides head pose angles into a number of non-overlapping ranges. During training,for each range of head pose angles,a facial landmark template is constructed by using the CPR (Cascaded Pose Regression) method,and a corresponding texture model is established by using the CLM (Constrained Local Model) method. During testing,given a new face image,all the templates are applied to it,each resulting in a set of facial landmarks. The associated texture models are then used to compute the fitness values of them,from which the one with the maximum fitness is chosen as the final result. Experiments on the FERET database with comparison to a state-of-the-art method prove the effectiveness of the proposed method in localizing the facial landmarks on face images with large pose angles.关键词
人脸识别/多视角/人脸特征点定位/模型选择Key words
face recognition/multi-view/facial landmark localization/model selection分类
信息技术与安全科学引用本文复制引用
石正权,赵启军,陈虎..基于CPR和CLM的多视角人脸特征点定位方法[J].计算机技术与发展,2015,25(11):1-5,5.基金项目
国家自然科学基金资助项目(61202160,61202161) （61202160,61202161）
科技部重大仪器专项(2013YQ49087904) （2013YQ49087904）
