一种基于块共同特征值的人脸识别方法

主成分分析与线性判别分析是人脸识别的重要识别方法,它们都通过求解特征值问题实现特征提取,但由于维数灾难会导致小样本和奇异性问题。提出了一种简单的人脸识别方法,无需进行奇异值分解,能有效地降低计算代价。首先将图像划分成块,然后计算多项式系数,得到友阵用于特征提取。基于两张不同图像的多项式系数友阵来计算对称阵。最后通过计算对称阵的零空间的零化度识别相似的人脸图像。为验证提出方法的有效性,在ORL、Yale和FERET人脸数据库上进行了实验。结果表明,该方法对于有较大姿态与光照变化的人脸识别具有较高的识别性能。

A face recognition method based on block common eigenvalues

The principal component analysis and linear discriminant analysis are important face recognition methods. Both of them can realize feature extraction by solving the eigenvalue problem. However, the small sample and singularity problem can be caused by the curse of dimensionality. We propose a simple face recognition method, which can effectively reduce computation cost without singular value decomposition. Firstly, we divide the image into blocks, and then calculate the polynomial coefficients to obtain the companion matrix which is used for feature extraction. A symmetric matrix based on the polynomial companion matrix of two different images is calculated. Finally, the nullity of symmetric matrix is calculated to recognize similar face images. Experiments on the ORL, Yale and FERET face databases verify the effectiveness of the proposed method. The results show that this method has high recognition performance in recognizing the face with big variation in pose and illumination.