摘要
针对传统的人脸识别算法在单训练样本的情况下识别率不佳的情况,提出一种结合拉普拉斯滤波与中心对称局部二值模式的人脸识别算法(LFCLBP)。对原始人脸图像进行拉普拉斯滤波处理;然后对图像提取梯度幅值和梯度相位信息,对梯度幅值用CS-LBP算子编码,再将梯度相位量化到16个区间进行编码,将二者融合成人脸图像的LFCLBP特征;分块统计直方图特征,将所有分块的直方图串联起来作为人脸图像的特征向量,并用最近邻分类器识别。在YALE人脸库和AR人脸库上进行测试,测试结果表明该算法有效,在光照变化、表情变化和部分遮挡等环境下对单样本人脸图像具有较好的识别效果。
To overcome the limitations of traditional face recognition methods for single sample,a novel method of facerecognition based on Laplace Filter and Center-symmetric Local Binary Pattern(LFCLBP)is proposed.Firstly,originalface images are filtered by Laplace filter.Secondly,gradient magnitude maps and phase maps of a face image are calculated.A operator named Center-Symmetric Local Binary Pattern(CS-LBP)is proposed to encode the gradient magnitude,andgradient phase is quantized into sixteen regions,then the proposed LFCLBP is the combination of the binary codes ofphase and CS-LBP of magnitude.Finally,LFCLBP feature maps are divided into several blocks,and the concatenated histogramcalculates over all blocks are utilized as the feature descriptor of face recognition.The recognition is performed byusing the nearest neighbor classifier.Experimental results on YALE and AR face databases validate that the LFCLBP algorithmis an outstanding method for single sample face recognition under different illumination conditions,different facialexpression conditions and partial occlusion conditions.
作者
杨恢先
刘凡
贺迪龙
YANG Huixian;LIU Fan;HE Dilong(School of Physics and Optoelectronics, Xiangtan University, Xiangtan, Hunan 411105, China)
出处
《计算机工程与应用》
CSCD
北大核心
2017年第7期165-170,共6页
Computer Engineering and Applications
基金
湖南省教育厅一般项目(No.13C931)
关键词
人脸识别
单样本
拉普拉斯滤波
中心对称局部二值模式
最近邻分类器
face recognition
single sample
Laplace filter
Center-Symmetric Local Binary Pattern(CS-LBP)
nearest neighbor classifier