Research on Binarization of Graphic Document Using Sparse Representations
Xia Kai-jia · Jisuanji fangzhen · 2013
In this paper,we proposed a denoising approach for document images or graphical images. Differ from the other denoising methods which have several drawbacks,the proposed approach used the direction information from graphics in the image. Initially,curvelete transform was applied on the image to generate a sparse representation of the noise image. Then by using basis pursuit denoising method,an optimal denoising image is estimated. In this framework,we also used the noise spread information to get the optimal solution. The experiment shows that the proposed method outperformances other traditional approach.