Check image binary with morphological method

Abstract

During the past few years, automatic check processing has become a popular topic in the field of document image analysis (DIA). Image binary is an important but hard task in automatic check image processing system. Difficulties in binary procedure derive mainly from the different types and positions of seal imprints. In this paper, we proposed a new binary method which is based on the fact that objects in a character image are mostly dark with thin strokes. In order to remove large block backgrounds, we firstly designed a set of morphological perorations to enhance the local feature of thin objects. Then a global threshold is selected and applied to the new enhanced image. Experiment results on 2745 real-life Chinese check images demonstrate the efficiency of our method compared with other commonly used binary methods. The recognition success rate has achieved 91.5%.