摘要
针对(2D)2PCA无法保存某些重要局部特征的问题,提出一种分块双向二维主成分分析融合局部特征方法。首先,将图像分解为互不重叠的子块,每个子块包含重要的局部信息,利用(2D)2PCA对子块进行特征提取并投影到特征子空间。然后,对每个子块分别设计一个分类器并在一定置信度范围内判别测试样本所属类别。最后,根据所有子块所属类别的置信度之和完成人脸分类。在四个人脸识别数据库上的实验结果表明,相比其他几种人脸识别算法,该方法取得了更高的识别精度。
We proposed a fusion method of blocked bidirectional 2DPCA and local feature because （2D）2PCA is unable to preserve some essential local features. First, the method decomposes the image into non-overlapping suh-blocks, each sub-block contains important local in- formation and is extracted its feature using （2D） 2 PCA＇and then projecting onto feature subspace. After that the method will design a classifier for every sub-block and identify within a certain confidence degree range the category of the test sample owned. Finally it achieves the faces classification according to the sum of the confidence degree of all the sub-blocks belonged. Experimental results on four face recognition data- bases showed that the proposed method achieved better recognition accuracy compared with several other face recognition algorithms.
出处
《计算机应用与软件》
CSCD
2015年第11期157-161,199,共6页
Computer Applications and Software
基金
国家自然科学基金项目(60527002)
关键词
人脸识别
双向二维主成分分析
特征提取
局部特征
置信度
Face recognition
Bidirectional 2DPCA
Feature extraction
Local feature
Confidence degree
