摘要
基于单样本的人脸识别具有重要的应用价值,然而对仅有一个注册样本的人脸图像进行识别是一个具有极大挑战性的问题。对近年来提出的单样本人脸识别的算法进行分类和介绍,以识别率为指标对比了这些算法的实验结果,同时给出了这些实验针对的人脸数据库、数据库的规模和训练/测试样本集的划分;总结了影响单样本人脸识别率的关键因素及各算法的优缺点,分析了一些算法取得较优识别率的原因及未来可能的研究方向。
Face recognition system based on one gallery sample is very valuable for less laborious effort for collecting images and lowering the cost for storing and processing them. However, it is very challengeable to correctly recognize a person from face database with only one sample for everybody. Some algorithms to deal with one sample problem have been proposed in recent years. They are re- viewed and introduced simply, The correct recognition rates in experiments of these algorithms are compared and the relevant issues such as database, class number, and how to divide training set and testing set are also discussed. Some key factors in one sample face recognition are pointed out and some promising directions for future research are also proposed.
出处
《西华大学学报（自然科学版）》
CAS
2014年第4期1-5,10,共6页
Journal of Xihua University:Natural Science Edition
基金
国家自然科学基金(61373163)
四川省教育厅资助科研项目资助(11ZB069)
四川省可视化计算与虚拟现实四川省重点实验室项目(PJ2012001)
关键词
人脸识别
单样本
特征提取
子空间学习
通用库
face recognition
single sample
feature extraction
subspace learning
generic database.. .~.
