Blurred license plate recognition via sparse representations

Blurred license plate recognition is one of the most challenging problems in computer vision. A novel license plate recognition framework based on no-negative sparse representation is proposed. In the proposed algorithm, the testing image is represented as a no-negative sparse linear combination of the dictionary, which is preprocessed from standard templates. The proposed framework include two stage: singe character estimation via no-negative block correlation and multiple-characters no-negative sparse representations optimization for refining the recognition results, which simultaneously fulfil the license plate de-blurring and recognition. The experimental results show that it can achieve a high recognition rate.