Abstract

This paper proposes a locality correlation discriminant with neighborhood preserving embedding for face recognition, which considers both the locality correlation and manifold structure of the training data. A new locality correlation preserving within-class scatter matrix is defined, which not only contains the locality preserving information but also contains the neighbor correlation information, and defines a novel objective function to learn the manifold structure of the data in a low-dimensional space. Since the obtained manifold structure takes consideration of the local neighbor correlation information and the discriminant information, it might be more accurate for characterizing the feature of face images. Experiments conducted on Yale and ORL database indicate the effectiveness of the proposed method.

Paper Details