Reduced Eigen Space Dimensionality for Fast Face Recognition

Principal Component Analysis (PCA) is one of the techniques to reduce dimensionality of images for face recognition. Two dimensional Principal Component Analysis (2DPCA) algorithm is used to reduce time complexity of PCA algorithm. In this paper, 2DPCA algorithm is modified and developed a new algorithm using transpose of extracted eigenvectors. The new algorithm is named as Two dimensional Principal Component Transpose Analysis (2DPCTA) algorithm. It extracts Principal components of two dimensional images and represents the face images using extracted eigenvectors as well as transpose of the selected eigenvectors whereas 2DPCA used extracted eigenvectors only to represent images. Therefore, the proposed technique reduces face space dimensionality as well as time complexity as compared to PCA and 2DPCA algorithms. The proposed algorithm has been implemented on various standard face databases e.g. Yale, ORL, Jaffe and KDEF face databases. The 2DPCTA, PCA and 2DPCA algorithms tested on faces databases independent of pose, expression and degradation variations. The results are analyzed with respect to time, space dimensionality and recognition rate in time domain as well as frequency domain. It is observed from experimental results that the proposed technique reduces face space dimensionality and time complexity including time of extracting eigenvectors, time of image representation as well as time of objects classification as compared to PCA and 2DPCA algorithms.