Based on sub-image feature combination, a feature extraction method of face image was proposed. Combined with BP neural network, a face recognition model was presented. In the model, face images were divided into sub-images firstly, then the discrete cosine transform was operated on each sub-image and the largest coefficient was selected to denote the sub-image, finally these coefficients were combined as the feature of the whole face image. BP neural network was used as classifier in the model and its parameters were optimized according to experiments. Based on ORL face database, experimental...