一种融合多模式韦伯局部特征的人脸识别方法  

A Face Recognition Method Based on Fusion Multi-modal Weber Local Feature

作　　者：李昆明[1] 王玲[1] 闫海停[1] 刘机福[1] 

机构地区：[1]湖南大学电气与信息工程学院,长沙410082

出　　处：《小型微型计算机系统》2014年第7期1651-1656,共6页Journal of Chinese Computer Systems

摘　　要：在分析韦伯局部算子的基础上,提出一种融合多模式韦伯局部特征的人脸识别方法.该方法先用韦伯算子提取人脸图像的差分激励和方向,对差分激励进行方向累积分解,在差分激励方向累积图像上用局部二值方法提取累积图像的特征,然后对方向进行差分求取韦伯方向差分二值编码,并串接差分激励特征和韦伯方向特征,然后用基于分块的线性判别进行降维,最后计算余弦相似度.在ORL和CAS-PEAL人脸库上,实验结果表明,该方法识别性能优于基于特征脸的人脸识别和基于Gabor滤波的人脸识别方法.该方法不但计算复杂度和空间复杂度大幅减少,而且能够有效提取图像的可区分特征,提升系统的识别性能.A recognition algorithm which is based on fusion multi-modal weber local features is proposed. Firstly,the algorithm extract weber difference excitation and orientation,and then decomposes and accumulates the difference excitation according to the quantization orientation,after that,codes the results using local binary pattern and extracts the histogram features. Secondly,extracts the weber orientation difference binary features and concatenates it to difference excitation features. Thirdly,reduces the dimension by block-based fisher linear discrimination. Lastly,calculates the cosine distance similarity. The experiences results obtained on the ORL and CAS-PEAL database shows that,the recognition ability of this method outperforms traditional eigenface-based algorithms and the gabor-based methods. The algorithm is not only reduces the time and space consume,but also can obtain the most powerful discriminate feature effectively.

关 键 词：韦伯局部算子 方向分解 累积 方向差分 特征融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]