摘要
对WLD特征进行改进。改进的特征提取方法为:首先将原始人脸图像划分为若干个子块,然后提取每块图像的WLD直方图统计特征,其中的梯度方向是用Prewitt算子计算的,最后将所有分块的WLD直方图序列连接起来构成特征向量。为了验证改进特征的性能,用支持向量机进行人脸识别,人脸图像取自ORL和YALE人脸数据库。实验结果表明,采用改进后的特征可以显著提高人脸识别率。
We improve the Weber local descriptor (WLD)feature.The improved feature extraction method is that to divide the original face image into a number of blocks first,then to extract the histogram statistical characteristics of WLD in each block,in which the gradient orientation is computed by Prewitt descriptor,and finally to concatenate the WLD histograms series of all blocks to form eigenvector.In order to verify the performance of the improved WLD feature,we use support vector machine (SVM)for face recognition,the face images come from ORL and YALE face databases.Experimental results show that to use the improved WLD feature can significantly enhance the face recognition accuracy.
出处
《计算机应用与软件》
CSCD
2015年第4期141-144,共4页
Computer Applications and Software
基金
国家自然科学基金项目(61170124)
关键词
人脸识别
韦伯局部描述符
支持向量机
直方图
Face recognition
Weber local descriptor
Support vector machine
Histogram