Binarization of Document Images with Complex Background

Thresholding a gray-level image into two levels is the first step and also a critical part in most document image analysis systems since any error in this stage will propagate to all later analysis. Various algorithms have been proposed over past years, however, with a complex background, problem remains unsolved. In this paper, we proposed a new document image binary method which is based on the gray level feature and the character stroke feature. We firstly designed a set of morphological perorations to describe strokes' local edge feature. Then a global threshold is selected based on a new projection histogram. Experiment results show the efficiency of our proposed method.