An image matrix compression based supervised locality preserving projections for face recognition

Recently, a new manifold learning algorithm named locality preserving projections (LPP) that aims at finding an embedding that preserves local information has been proposed and used for face recognition. In this paper, an image matrix compression based supervised locality preserving projections is proposed for face representation and recognition. In this new scheme, a bilateral-projection-based 2DPCA (B2DPCA) for image matrix compression is performed before supervised locality preserving projections. The bilateral-projection-based DPCA algorithm is used to obtain the meaningful low dimensional structure of the data space in this new method. Experiments based on the ORL face database demonstrate the effectiveness and efficiency of the new. Results show that the new algorithm outperforms the Laplacian faces which uses the locality preserving projections (LPP) and achieve a much higher accurate recognition rate.