Abstract: In order to improve the recognition rate of the face recognition, a face recognition method of modular Two-dimensional Supervised Local similarity and Diversity Projection(2DSLDP) is proposed in this paper. The original images are divided into modular images. Using similarity scatter and diversity scatter, it can compute the feature scatter. The modular images are combined according to a certain order to extract the features. The dimension of the original images can be depressed. The algorithm not only can reduce the relaxing of computer, but also can keep the local features of the images effectively. Experimental results indicate that the proposed method has a higher recognition rate.

Key words: face recognition, 2DSLDP algorithm, feature extraction, modular matrix

摘要： 为提高人脸识别算法的鲁棒性，提出一种有监督的二维分块局部相似与差异的人脸识别算法。该算法对原图像矩阵分块后，利用局部相似和差异算法中定义的2个权值矩阵，求解分块矩阵中的投影矩阵，将得到的投影矩阵按次序整合得出特征矩阵，以达到使原图像降维的目的。实验结果表明，该算法在降低计算难度的同时，能保持图像的局部信息，取得良好的识别效果。

关键词: 人脸识别, 2DSLDP算法, 特征提取, 分块矩阵

CLC Number: 

• TP391.4