A New Method to Extract Face Features Based on Combination of Mean Face SVD and KFDA

A new method to extract features of face image based on Singular Value Decomposition (SVD) and Kernel Linear Discriminant Analysis(KFDA) is proposed. First, the mean image of all train samples is selected as a standard face image, and all the train samples are projected into the two orthogonal matrixes which come form the SVD of the mean face image. Then the left-top elements of projecting coefficient matrix is extracted as the primary features. Finally, KFDA is used to extract the recognition feature. In this method, the problem of the SVD used into face recognition is resolved, at the same time, label information of train samples is considered and non-linear feature is also extracted. Experiments are done on ORL and CAS-PEAL databases, the results show the method is effective.