Automatic face recognition is a challenging problem in the biometrics area, where the small sample size problem exists. An Enhanced Relation Discriminant Analysis (ERDA) method is proposed to solve the small sample size problem. In our framework, the neighbor and class relations of data are used to construct the embedding for classification problems. The proposed algorithm learns the embedding for the submanifold of each class by solving an optimization problem. After being embedded into a low-dimensional subspace, data points tend to move due to local intra-class attraction or inter-class repulsion. ERDA aims to map the image space into a submanifold that faithfully discovers the local discriminative manifold structure of face image. This method accounts for both the representation and the classification points of views. Experimental results on the AT&T and Yale face image databases demonstrate the effectiveness of the method.