A Novel Subspace Method for Face Recognition

Feature extraction is the key problem for face recognition. Many methods have been proposed, and among these methods the subspace method has been given more and more attention owing to its good performance. In this paper, a novel subspace method called Inverse Fisher discriminant with Schur decomposition (IFDS) is proposed for face recognition. In comparison with Inverse Fisher discriminant analysis (IFDA), IFDS eliminates linear dependences among discriminant vectors. Experiments results on ORL and FERET face database demonstrate that IFDS outperforms Fisher discrimiant analysis (FDA) and IFDA algorithm.