In order to solve the problem that classifiers do not make full use of the differences between different types of face samples in face recognition, an effective method for face recognition was proposed, namely Complement Null-Space (CNS) algorithm; and further more, another method which combined CNS and nearest space Distance (CNSD) was proposed. Firstly, subspaces and complement null-spaces of all types of training images were constructed. Secondly, the distances between the test image and all types of subspaces as well as the distances between the test image and all types of complement null-spaces were calculated. Finally, the test image was classified into the type which has the minimum subspace distance and the maximum complement null-space distance. On ORL and AR face databases, the recognition rates of CNS and CNSD are much higher than those of Nearest Neighbor (NN), Nearest Space (NS) method and Nearest-Farthest Subspace (NFS) method when the number of training samples is small; and it is a little higher than that of NN, NS and NFS when dealing with large samples. Simulation results show that the proposed algorithm can make full use of the differences between different types of images and has good recognition ability.