Hill climbing algorithm for license plate recognition

Histogram thresholding has been widely used for image processing - it is simple, fast, and computationally inexpensive. In this paper, we develop a creative approach based on histogram's distributions to segment interest regions from background. Unlike the existing threshold detection methods which measure the statistics of histogram in the multi-modal images, our approach analyses the shape representation of multi-modal which has several hill-climbing curves. The behavior of algorithm works like human vision which focuses on the high contrast areas and scans the shape variation first. Moreover, such an algorithm presents a new type of histogram analysis that depends on the particular shape of certain distribution in histogram. Experimental results reveal that the proposed algorithm performs distinct effects especially on the recognition of artificial signs such as road sign, vehicle plate, and signboard.