On effectiveness of image analysis and recognition by principal component method and linear discriminant analysis

In paper, some aspects of image analysis based on principal component analysis and linear discriminant fnalysis are considered. The image recognition technique on base this methods consists of two steps: first we project the face image from the original vector space to a reduced subspace of principal components, second we use LDA to obtain a linear classifier. Main attention is focused on the development of efficient algorithm for computing principal components for large image set. A linear condensation method is used as a new technique to calculate the principal components of a large matrix. To improve the efficiency of the linear condensation method is proposed to use a process of block diagonalization of the matrix. The accuracy and high performance of the developed algorithm is evaluated.