The Hermite-feature-based kernel discriminant analysis(KDA) method for face recognition is presented.Firstly,a multiresolution-directional-oriented Gabor-like Hermite transform is used to derives the facial features that are characterized by spatial frequency,spatial locality,and orientation selectivity to do with the variations due to illumination and facial expression changes.Then the KDA method is used to enhance the face recognition performance.The KDA can extract the features particularly useful in discriminating ability in the high-dimensional space,which are the most discriminating nonlinear features in the input space.It has been demonstrated to be capable of nonlinear discriminant analysis.The feasibility of the method has been successfully tested.