分块二维最大间距准则的特征提取方法

Modular 2DMMC for Feature Extraction

针对二维最大间距准则(Two Dimensional Maximum Margin Criterion,2DMMC)算法进行特征提取时,无法提取局部的特征.同时,该算法也受不同的表情、光照以及姿态等条件的影响,识别的效果也大大降低.因此,提出一种基于分块二维MMC(Modular Two Dimensional Maximum Margin Criterion,M2DMMC)的人脸识别方法.首先,对图像矩阵进行分块,然后对分块后的矩阵进行2DMMC特征抽取,对每一子块抽取的特征进行整体融合,最后采用最近邻判决准则进行分类识别.在ORL,Yale人脸图像库进行实验的结果表明,新算法相对于MMC算法、二维MMC算法以及分块MMC算法均有较好的识别性能.

To tackle Two-Dimensional Maximum Margin Criterion（ 2DMMC ） algorithm for feature extraction, which can not effective- ly extracts local features. Meanwhile, the algorithm was affected by different facial expressions, illumination and pose, etc. So, Modular Two Dimensional Maximum Margin Criterion （ M2DMMC ） is proposed in this paper. First, in proposed approach, the original images are divided into modular images, which are also called sub-images. Then, M2DMMC method is directly used to the sub-images to ob- tain feature from the previous step. Features of sub-images are combined into global Features. At last, the recognition results are ob- tained by Nearest Neighbor （NN） classifier. By the test on ORL,and Yale face database,the results show that the proposed algorithm with respect to the MMC,2DMMC and MMMC algorithm has better recognition performance.