Removal of speckle and impulse noise from Indian License Plates using a 2-Stage Median Filter

In License Plate Recognition algorithms the extraction of the license plate region accurately is the seed and lifeline of the entire LPR system. It consists of a sequence of pre-processing algorithms to remove the bottlenecks like, presence of non uniform outdoor illumination conditions during image acquisition, motion blur due to varying vehicle speed and changing dynamic backgrounds. The acquired images are subject to image corruption due to noise. This paper presents an algorithm namely 2-Stage Median Filter to de-noise the images corrupted by fixed value impulse noise in Indian License Plates. First stage is detection of impulse pixel depends on threshold values and second stage is filtering of corrupted pixels which are replaced by the median of uncorrupted pixels in the filtering window. Experimental results show that the proposed algorithms provide a flexible and reliable way in de-noising the highly corrupted image due to dynamic environmental condition of License plate capturing.