Abstract:
Character recognitions are generally very sensitive to skew in the automatic license plate recognition (ALPR) system. The skew correction of license plate thus is an important step in ALPR. In this paper, we propose an orientation-based method for skew detection. The license plate image is firstly divided into a set of 5x5 non-overlapping blocks. The local orientation of each black is estimated by gradients of pixels in the block. Next, the direction histogram which can reveal the overall orientation information in the license plate image is counted. The skew angle of license plate is detected by the local maximum of the direction histogram. The experimental results demonstrate the great robustness and efficiency of our method.
