Abstract: Many environmental factors and improper handling cause ancient document to suffer a high degree of degradation. In order to solve the problems that arise in the binarization of old books images having a uniform character-size characterized by variations in image contrast, we try to design a binarization algorithm in order to better reveal the contents recorded in ancient books. Fist we will detect text stroke edges by Canny's method and calculate stroke widths, the ancient books image background is estimated through the calculated stroke width. Next step, we compensate the ancient document contrast by using the estimated background. Then, the document text is segmented by Howe's binarization algorithm. Further improvements are achieved by a postprocessing step. Finally the performance of the algorithm is analyzed and evaluated, the proposed method is better than Mesquita' method which achieved the best performance in the ICFHR'2014.