基于改进AAM的人脸特征点提取  

Face feature points extraction based on refined AAM

作　　者：牛星[1] 席志红[1] 金子正秀 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001 [2]东京电气通信大学情报理工研究科

出　　处：《应用科技》2011年第4期35-38,共4页Applied Science and Technology

摘　　要：提出了具有鲁棒性的基于局部二值模式LBP纹理提取人脸特征点的主动外观模型AAM算法(LBP-AAM).首先建立3种人脸模型实例(正面,左转及右转人脸模型);然后应用LBP判断和预测测试图片的旋转类型,依据预测结果选择合理的模型实例去匹配;最后提取出人脸特征点.实验结果证明本方法比传统的AAM方法在精度上提高了27%,效率上提高了9%.In this paper,we propose a robust facial feature points extraction method-LBP-AAM（L-AAM）,using active appearance model（AAM） based on local binary pattern（LBP） texture features.We firstly generated three types of model instances（frontal,left-rotated and right-rotated）,and LBP was used to judge the type of test facial image and predict the rotation of the test the face,and according to the prediction we selected proper model instances as the fitting model.Finally we extracted the feature points of the face.Experimental results proved that this method increased the fitting accuracy rate by about 27% and the time consumption was decreased by about 9% comparing with the standard AAM method.

关 键 词：主动外观模型 局部二值模式 局部二值模式主动外观模型 模型实例 人脸特征点 

分 类 号：TP232[自动化与计算机技术—检测技术与自动化装置]