International Journal of Intelligent Engineering and Systems

10

Authors: Xiaohua Zhang, Yuelan Xin, Ning Xie, Wenbo Jiang, Yue Zhao

As a pre-processing step for document analysis, binarizing document image is often performed. The binarization is regarded as a necessary step to separate text and background. When document images are acquired under poor lighting circumstances, the images are soiled with uneven shading. This paper presents an improved approach for binarizing document images which are illuminated unevenly. We employ 2D polynomials with order 3 to fit very complicated uneven shading surface of document images. The document image is divided into several blocks overlapped partially with each other to reduce the discontinuity between polynomials. To improve the estimation accuracy, we normalize pixel coordinates to establish a stable equation system for computing the polynomial coefficients. After removing the uneven shading, most part of the background is filled with the maximum of surface and then a global thresholding is applied. In the experiment section, we demonstrate that the binarization results are satisfied, compared with other methods.