基于小波变换和多特征融合算法的人脸识别

计算机工程与应用 ›› 2016, Vol. 52 ›› Issue (12): 201-204.

基于小波变换和多特征融合算法的人脸识别

关学忠，王文锋，张新城，尹廷武，张璐

东北石油大学电气信息工程学院，黑龙江大庆 163318

出版日期:2016-06-15 发布日期:2016-06-14

Face recognition method based on wavelet transform and multi-feature fusion algorithm

GUAN Xuezhong, WANG Wenfeng, ZHANG Xincheng, YIN Tingwu, ZHANG Lu

School of Electrical Information Engineering, Northeast Petroleum University, Daqing, Heilongjiang 163318, China

Online:2016-06-15 Published:2016-06-14

摘要/Abstract

摘要： 提出了基于小波变换和多特征融合算法的人脸识别方法。该方法先对原始人脸图像进行简单加权小波变换以降低维数，施行改进的模块二维主成分分析（M2DPCA）抽取特征，再进行加权最大散度差鉴别分析（WMSD）得到最终的特征图像，采用最近邻分类器对人脸分类识别。该方法不仅利用了人脸图像的局部特征和类别信息，而且避免了矩阵的奇异值分解可能遇到的问题。在ORL人脸库上实验，以验证该方法的有效性。

关键词: 简单加权小波变换, 模块二维主成分分析（M2DPCA）, 加权最大散度差鉴别分析（WMSD）

Abstract: A face recognition method based on wavelet transform and multi-feature fusion algorithm is proposed. Firstly the weight wavelet transform is used to the original face images for dimension reduction. Then the improved Modular 2 Dimensional Principal Component Analysis（M2DPCA） is applied to the images for feature extraction, and then the Weight Maximum Scatter Difference discriminate analysis（WMSD） is used to the sub-images of these obtained feature images in which way the final feature images are obtained. Finally, the nearest neighbor classifier can distinguish the different human face. This method can not only exploit local feature of face and discriminate information but also totally avoid the problem of singular value decomposition of matrix. Experiments performed on ORL face database verify the effectiveness of the proposed method.

Key words: weight wavelet transform, Modular 2 Dimensional Principal Component Analysis（M2DPCA）, Weight Maximum Scatter Difference discriminate analysis（WMSD）

关学忠，王文锋，张新城，尹廷武，张璐. 基于小波变换和多特征融合算法的人脸识别[J]. 计算机工程与应用, 2016, 52(12): 201-204.

GUAN Xuezhong, WANG Wenfeng, ZHANG Xincheng, YIN Tingwu, ZHANG Lu. Face recognition method based on wavelet transform and multi-feature fusion algorithm[J]. Computer Engineering and Applications, 2016, 52(12): 201-204.