计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (10): 163-165.

• 图形、图像处理 • 上一篇    下一篇

基于改进的ASM方法的人脸特征点检测

孙延鹏1,傅 蓉2   

  1. 1.青岛大学,山东 青岛 266071
    2.上海中科院 计算技术研究所,上海 201203
  • 收稿日期:2007-09-25 修回日期:2007-12-27 出版日期:2008-04-01 发布日期:2008-04-01
  • 通讯作者: 孙延鹏

Face feature points detection based on refined ASM

SUN Yan-peng1,FU Rong2   

  1. 1.Qingdao University,Qingdao,Shandong 266071,China
    2.Shanghai Institute of Computing Technology,Chinese Academy of Sciences,Shanghai 201203,China
  • Received:2007-09-25 Revised:2007-12-27 Online:2008-04-01 Published:2008-04-01
  • Contact: SUN Yan-peng

摘要: 针对传统的Active Shape Model(ASM)方法的缺陷与不足,提出了几点改进的方法。传统的ASM方法利用人脸检测提供的初始位置进行特征搜索;在改进算法中,利用了瞳孔定位的方法进行初始化,并采取了边缘约束,自适应步长等改进方法,有效地改善了ASM方法的性能。实验结果表明,改进的ASM方法在准确性和鲁棒性上有较大提高。

关键词: 主动形状模型, 瞳孔定位, 边缘约束, 自适应步长, 人脸识别

Abstract: In view of the traditional Active Shape Model(ASM) method’s flaw and insufficiency,the paper proposes several improvements methods.The traditional ASM method carries on the characteristic search using of the initial point which has provided using the person face examination;in the improved algorithm,the paper has used the method of pupil orientation on the initialization and adopted improvement methods of edge restraint,auto-adapted length of stride to effectively improved the ASM method performance.The experimental result indicates that,the improved ASM method has a bigger enhancement in the accuracy and robustness.

Key words: Active Shape Model(ASM), pupil orientation, edge restraint, auto-adapted length of stride, face recognition