Extraction of Legal Amount from Color Bank Check Images

In this paper, a novel method to extract legal amount from color Chinese bank check images for recognition task is presented. Firstly, global binarization method is applied to each channel of color legal amount images to get multiple binarization of source images, and get coarse binarization result through combining multiple binarization results. Then, a new method based on writing-rule of legal amount is proposed to locate text area and remove noise. Finally, local binarization method is applied to text area to refine result. Experimental results are given in the paper to demonstrate effectiveness of the proposed method.