Two-dimensional neighborhood preserving embedding for face recognition

Neighborhood Preserving Embedding (NPE) is a subspace learning algorithm, which has the ability of preserving local neighborhood structure on the data manifold. Though NPE has been applied in many domains of pattern recognition, it is a vector-based method and will be encountered the small size sample (SSS) problem when it is directly applied to face recognition. To address this problem, the popular method is to use PCA prior to performing NPE, but the pre-processing procedure using PCA could result in the loss of some important discriminatory information. In this paper, a novel method called two-dimensional neighborhood preserving embedding (2DNPE) is proposed to extract the features for face recognition. Extensive experiments are performed to test and evaluate the new method using ORL and Yale face database. The experimental results indicate that the 2DNPE method has better face recognition performance and more effective.