Binarization of Degraded Document Images Based on Combination of Contrast Images

This paper introduces a novel algorithm for binarization of gray scale images of degraded documents based on the creation of three different structural contrast images, the representation of these features in a 2D feature space that is further partitioned. A weak bi-level image and a strong bi-level image are created and combined to produce the final image. Experiments were conducted with the application of the new algorithm to DIBCO 2011 and H-DIBCO 2012 datasets achieving satisfactory results.