A New Fractal Series Expansion Based Enhancement Model for License Plate Recognition
Pinaki Nath Chowdhury,Palaiahnakote Shivakumara,Hamid A. Jalab,Rabha W. Ibrahim,Umapada Pal,Tong Lu
DOI: https://doi.org/10.1016/j.image.2020.115958
2020-01-01
Abstract:License plate recognition is an emerging topic for real-time applications in smart city development because of automatic systems for toll fee-paying, traffic controlling, and vehicle detection. Speed of vehicles, unpredictable weather conditions such as night/low light/limited light images, and capturing images at different angles make recognition harder. This paper presents a new Fractal Series Expansion (FSE) model for license plate image enhancement. The proposed FSE model is justified because it estimates the high probability for the pixels which represent the license plate compared to the pixels which represent background irrespective of the above challenges, resulting in enhanced image. Besides, since the FSE model considers local information for estimating probability, the model has the ability to tackle non-uniform degradations as well as distortion affected multiple adverse factors. In addition to qualitative results, to validate the effectiveness of the proposed enhancement, quantitatively, recognition rates of the different methods before and after enhancement are computed. For this purpose, we have considered different datasets like dataset of Night License Plate Images (NLPI), which consists of images captured in the night and low lights environment, the UCSD benchmark dataset which provides poor and high quality day license plate images, etc. It is noted that recognition results after enhancement is higher than that of before enhancement, and hence our enhancement is useful and effective.