One of the key issues of face recognition is to extract the features of face images, and a new method of feature extraction, two-dimensional discriminant locality preserving projections (2D-DLPP), is proposed. 2D-DLPP benefits from three techniques, i.e., locality preserving projections (LPP), image based projection and discriminant analysis. Firstly, LPP is an effective feature extraction method that optimally preserves the local structure of the samples. Secondly, compared to vector based projection, image based projection can reduce the complexity of algorithm, avoid the small sample size problem and give more spatial structural information of image. Finally, discriminant analysis applied in 2D-DLPP can improve recognition performance by maximizing the interpersonal distance and minimizing the intrapersonal distance. Experiments are performed to test and evaluate the algorithm using the ORL and the Yale face databases. The Experimental results show that 2D-DLPP has better face recognition performance than other methods.