Macau License Plate Character Segmentation Through ROI detection and Redundant Region Removal Method

License plate character segmentation links the function between license plate detection module and character recognition module. Variable-length license plate character segmentation is a challenging task due to the variations of license plate styles. In this paper, our work is focused on the character segmentation task. It is based on the detection results of “Region of Interest(ROI)” detection method. The main contributions include two parts. Firstly, a reference region is estimated according to candidate regions by median statistics of width and height. The reference region is used to compare with all the candidate regions, aiming to remove false positives. Secondly, a redundant region removal method is proposed. It is implemented by removing cross regions which are located at the same location of character. Experimental results on Macau license plates show that the proposed method achieves promising results with 99.37 % segmentation accuracy.