Abstract: According to the low chromatic difference and poor image quality characteristics of the pressed protuberant characters, the Eigenfaces method for the overall feature extraction and recognition is adopted. However, without considering the between-class variance, the Eigenfaces method can’t achieve ideal classifying results. A novel method based on the reconstruction errors for the character recognition is proposed. Firstly, a subspace is established for every class. Then, images of the test samples are reconstructed under every subspace. Finally, the recognition experiment is carried out according to the root mean square error between the original image and the reconstruction image. The proposed method makes full use of the features of the within-class and the variance of the between-class. The experiment results show that the method not only remarkably improves the recognition rate but also meets the real-time requirement.

Key words: Eigenfaces, Protuberant characters, Reconstruction error, Subspace