In this paper, we propose a method based on classified bases and example-based super-resolution to reconstruct the high-resolution details of a single low-resolution car license-plate image. The proposed method first uses the Bayesian classifier to classify individual characters of a license plate in the Orthogonal Locality Preserving Projections (OLPP) subspace. The proposed cost function in the refinement step then corrects the initial classification result to further improve the classification accuracy. Based on the refined classification result, in the final class decision step, the best basis set is determined to reconstruct the high-resolution image. Our experimental results show that the proposed method effectively improves the visual quality of reconstructed high-resolution license plate images.