PROJECT TITLE :

Bi-level thresholding for binarisation of handwritten and printed documents

ABSTRACT:

Document image binarisation algorithms have been available in the literature for decades. However, most of the state-of-the-art methods address specific image degradation or characteristics. Moreover, they require one or more parameters to be tuned manually so as to present a significant binary image. In this study, a hybrid approach for document binarisation is presented. In the pre-processing stage, the degradation in the background image is smoothed using the L0-gradient minimisation algorithm and the foreground is enhanced using the local contrast feature. A divide and conquer based recursive auto-thresholding algorithm is then utilised to binarise the enhanced image. The proposed algorithm is evaluated objectively using the evaluation metrics such as F-measure, peak signal-to-noise ratio, negative rate metric. The extensive experiments over the different datasets including the Document Image Binarization Contest (DIBCO) 2009, Handwritten Document Image Binarization Competition (H-DIBCO) 2010, DIBCO 2011 and H-DIBCO 2012 show that the proposed hybrid binarisation algorithm outperforms most of the state-of-the-art algorithms significantly.