A Novel Approach for Automated Skew Correction of Vehicle Number Plate Using Principal Component Analysis

The performance of vehicle number plate recognition system is adversely affected by poor quality of captured images, especially, off-angled captured images (Skew images). This paper presents a novel method for vehicle number plate skew correction to improve the recognition rate. Firstly, the license plate is localized using pre-processing algorithm and wavelet decomposition and then proposed method of skew correction is applied. Principal component analysis (PCA) is used to find out the accurate skew (slope of the inclination) in both, horizontal and vertical directions and accordingly it has been accurately corrected. Adequate experimentation has been carried out on various types of skewed images and results have been compared with existing methods. Experimental results and quantitative comparisons reveal an effectiveness of the proposed method over existing methods of skew correction. The proposed method is not only capable of accurate skew correction of all possible types of skew but computationally efficient also.