[Paper] Using Gaussian Kernels to Remove Uneven Shading from a Document Image

抄録

Segmentation of document images into text or drawings is an important process, which is often related to binarization of a document image to perform character recognition and document analysis. This process is easier to do using a document image with a uniform background and illuminated under well-conditioned lighting. However, when a document image is very unevenly shaded, binarization becomes very difficult or even impossible. An effective method is to remove the shading prior to binarization. In this paper, we propose a novel method for estimating an unevenly shaded surface in an image obtained under poor illumination. A one-dimensional Gaussian kernel model is applied in both the horizontal and vertical directions to estimate the background surface, allowing uneven shading to be removed from the document image. Thereafter, the image can easily be binarized. Results of experiments conducted on many document images demonstrate that our method yields better results than other methods.