基于改进的ASM人脸特征点定位  

Facial Feature Extraction Based on Modified Active Shape Model

作　　者：邱丽梅[1] 

机构地区：[1]三明学院物理与机电工程系,福建三明365004

出　　处：《三明学院学报》2010年第2期124-127,共4页Journal of Sanming University

基　　金：福建省教育厅科技项目(JK2009037);三明学院教改项目(L1017/Q)

摘　　要：针对传统ASM算法的初始模型不准确会导致特征点提取失败,初始模型不仅复杂,而且不能适应多姿态人脸的特征点定位及收敛速度慢的缺点,提出了引入仿射变换不变性初始化原则(ATIIA)来建立初始模型。该模型使ASM算法不仅能快速、准确地提取人脸特征点,而且还能适应多姿态的人脸图像特征点的提取。The shortcomings of traditional ASM algorithm were that inaccurate initial model would lead to feature extraction failures,and initial model was not only complex,but could not meet the feature location of the multi-view face and slow convergence.In allusion to these shortcomings,this paper proposed an affine transform insensitive initialization algorithm to build the initialization model,which enabled the ASM algorithm not only extract facial features fast and accurately,but also adapt to feature extraction of multi-view face images.

关 键 词：初始模型 仿射变换 活动形状模型 人脸特征点定位 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]