Binarization of noisy gray-scale character images by thin line modeling

In this paper, we propose two new methods for the binarization of noisy gra y-scale character images obtained in an industrial setting. These methods a re different from other conventional binarization methods in that they are specially designed to detect only character-like regions. They exploit the fact that characters are usually composed of thin lines (strokes) of unifor m width. We first model the shape of the cross section of a character strok e and discuss how to detect the character stroke. Then, ALGORITHM I, which is a direct realization of our basic idea, is introduced, followed by an ad vanced algorithm named ALGORITHM II. The key to these algorithms is the loc al binarization-voting procedure. The performance of our methods is evaluat ed and compared with that of five other binarization methods using 550 slab ID number images, where a common character segmentation routine is attache d to each of the different binarization routines and the segmentation succe ss rate for each method is obtained. Experimental results show that ALGORIT HM II results in far better performance than the other methods. (C) 1999 Pa ttern Recognition Society. Published by Elsevier Science Ltd. All rights re served.