Shearlet feature manifold for face recognition

Face recognition is one of the major challenging problems in image processing. To effectively deal with this issue, a novel Shearlet feature manifold method for face recognition is introduced in this paper. Specially, the Shearlet feature is first extracted to capture the geometry and edge structures of face image; then the obtained high-dimensional Shearlet features are projected into low-dimensional subspace by using local geometry analysis which can simultaneously consider intraclass geometry and interclass discrimination information; finally, the face recognition is realized in the feature space by using the nearest-neighbor classifier. The experimental results on two face datasets show the effectiveness of this proposed algorithm.