| 摘要: |
| 提出了在不同环境下视频序列中人脸检测定位的多级算法.首先,用改进的帧间差自适应阈值法检测运动区域.然后,用YCbCr颜色空间检测肤色区域,并用二值形态学操作进行后处理.最后,构造眼嘴映射检测定位人脸.实验结果证明,该算法可以有效运用于多人脸、不同表情和姿态、不同光照的情况,检测率平均达到81%. |
| 关键词: 运动分割 肤色分割 特征验证 人脸检测 |
| DOI： |
| 分类号: |
| 基金项目:2013浙江省可视媒体智能处理技术研究重点实验室开放基金项目(2013010) |
|
| Face detection and location in video sequences |
|
YU Yulian1, LI Li2
|
|
1.Department of Communication Engineering,Shanghai Technical Institute of Electronics & Information;2.College of Information,Mechanical and Electrical Enginnering,Shanghai Normal University
|
| Abstract: |
| This paper presents a multilevel algorithm of human face detection and location in video sequences under different environments.First,it applies the method of improved inter-frame difference self-adaptive threshold to detect motion region.Then,it checks skin-like areas in YCbCr color space,and uses binary morphological operations for post-processing.Finally,it constructs eye and mouth maps to detect and locate faces.Experimental results show that the algorithm can be effectively applied to detect plural faces with different expressions and gestures under different lighting situations.The detection rate is up to 81% on average. |
| Key words: motion prediction skin color segmentation feature verification face detection |
