首页 | 本学科首页   官方微博 | 高级检索  
     

CPEF:一种用于人脸识别的图像择优方法
引用本文:王迪,陈岳林,蔡晓东,甘凯今.CPEF:一种用于人脸识别的图像择优方法[J].电视技术,2016,40(7):113-117.
作者姓名:王迪  陈岳林  蔡晓东  甘凯今
作者单位:桂林电子科技大学机械工程学院,广西桂林,541004
基金项目:2014年国家科技支撑计划课题 2014BAK11B02. 2014年广西科学研究与技术开发计划项目 桂科攻14122007-5;
摘    要:针对人脸识别的输入图像信息量高低不均的问题,提出了一种多参数图像择优的评分机制CPEF主要有三点贡献:第一点是提出了一种新颖的基于主动形状模型(ASM)定位的人脸水平旋转程度评估方法;第二点是采用递归的方式计算权重;第三点是提出了一组参数定量地描述了图像择优方法.过程如下:首先,计算图像质量;其次基于二分类器判断人眼是否处于睁开状态;然后以颜色直方图信息判定人是否处于张嘴状态;最后,综合水平旋转程度及其他参数,配合权重,给出评分.在FERET人脸库上的实验结果证明,CPEF有效地将人脸图像按照入主观感受做出排名,准确率高达92%.并且按得分高低排序采用Gabor配合SVM做人脸识别对比测试,对比实验证明,经CPEF处理后的图像在人脸识别中的准确率显著提升.

关 键 词:ASM  CPEF  旋转  递归
收稿时间:2015/9/10 0:00:00
修稿时间:2015/12/10 0:00:00

A evaluation method based on CPEF for face recognition
Wang Di,Chen Yuelin,Cai Xiaodong and Gan Kaijin.A evaluation method based on CPEF for face recognition[J].Tv Engineering,2016,40(7):113-117.
Authors:Wang Di  Chen Yuelin  Cai Xiaodong and Gan Kaijin
Affiliation:School of information and communication,Guilin University of Electronic Technology,School of information and communication,Guilin University of Electronic Technology,School of information and communication,Guilin University of Electronic Technology,School of information and communication,Guilin University of Electronic Technology
Abstract:This paper presents a novel mechanism named CPEF(Composite Parameters for Evaluating Face images) for selecting an image to conduct the best result for face recognition applications from an image sequence. Firstly, an method based on ASM positioning for the evaluation of image rotation is provided. Then a recursive scheme for parameters weighing is utilized. Finally, a composite parameter model is proposed to rate the appropriate degree of an image for face recognition. The experimental results show that the proposed method produces similar results to subjective evaluation and the accuracy is 92%. Furthermore, the effectiveness of this work is proved by face recognition applications using Gabor feature and SVM classifier.
Keywords:ASM  CPEF  Rotate  Recursive Function
本文献已被 万方数据 等数据库收录!
点击此处可从《电视技术》浏览原始摘要信息
点击此处可从《电视技术》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号