Abstract: This paper presents improved thresholding method called “Multi-Scale Adaptive NICK thresholding method” for Automatic License Plate Recognition (ALPR) System. By applying Connected Components Labelling (CCL) algorithm on the Binary image, license plate characters candidates will be determined. Improved thresholding method is vital to approach all the license plate characters on the binary image otherwise we might lose license plate characters under various illumination conditions and different weather conditions like rain, snow and fog. Most advantages of this method in compare with Niblack and Savola thresholding method is that it could be adapted for both types of degradation: interfering degradation and intensity degradation depends based on scale or window size, therefore you can improve all types of degradation. The interfering degradation will be restored on high scales size to cover sufficient data from the original input imag. In compare, intensity degradation needs very small scales to deal with disjoint and broken characters in the license plate. In this method we don’t miss any object in the binary image as license plate character’s candidate.

Keywords: ALPR (Automatic License Plate Recognition) system, NICK thresholding method, Connected Components Labelling (CCL), Savola method, Niblack method.