A New Binarization Method For Degraded Printed Document Images.

Abstract

In this work, we propose a new method to binarize the degraded printed images. This binarization use the texture information calculated from the edge image to enhance the binarization results for degraded documents. As pre-prossing stage, the wavelet transform followed by the wiener filter were used. Then, texture information has been extracted from the edge image. After, the extracted information will be combined with the texture features computed from the pre-processed image, to estimate the binarization threshold using Niblack’s method. Experiments are conducted on a set of document image with different types of degradation and DIBCO Datasets, the results were compared with three binarisation methods which are: Sauvola, Niblack and Rani’s methods. The results affirm that our binarization technique provides good results in existence of different deterioration as compared with other methods.