多模态人脸识别中的人脸特征定位  

FEATURE POINT DETECTION BASED ON MULTI-MODAL FACE RECOGNITION

作　　者：刘晓宁[1] 周明全[1] 耿国华[1] 

机构地区：[1]西北大学计算机科学系可视化研究所,陕西西安710069

出　　处：《计算机应用与软件》2007年第4期13-15,共3页Computer Applications and Software

基　　金：国家自然科学基金项目(No.60271032)。

摘　　要：多模态人脸识别技术是解决目前二维人脸识别领域瓶颈问题的一项重要手段。在二维人脸特征定位中结合Gabor变换和人脸弹性图方法,采用缩小搜索范围的方法提高系统效率;三维人脸特征定位首先采用PS方法进行粗略定位,然后仿照二维人脸识别方法定义Gabor变换系数向量进行精确定位,其中也采用一个模板来缩小搜索范围。试验证明,方法不仅效率高,而且能适应多种姿态和表情的变换。Multi-modal face recognition is an important way to solve the problem of 2D face recgnition. Gabor wavelet transform and Face Bunch Graph are integrated to locate the 2D face feature points, to improve the efficiency, a model of 11 × 11 is used to dwindle the range data. When to locate the 3D face feature points ,Point Signature is used to locate coarsely ,then Gabor is adopted to locate precisely. A model of 25 × 25 is also used to dwindle the range. Experimental results show that presented methods perform well and adaptive to pose and expression.

关 键 词：人脸识别 GABOR变换 人脸弹性图 多模态 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]