PROJECT TITLE :

Phase Based Binarization of Ancient Document Images Model and Applications - 2014

ABSTRACT:

During this paper, a part-primarily based binarization model for ancient document images is proposed, along with a postprocessing method that can improve any binarization methodology and a ground truth generation tool. Three feature maps derived from the part data of an input document image constitute the core of this binarization model. These options are the maximum moment of section congruency covariance, a locally weighted mean section angle, and a phase preserved denoised image. The proposed model consists of 3 customary steps: 1) preprocessing; two) main binarization; and three) postprocessing. Within the preprocessing and main binarization steps, the features used are mainly phase derived, whereas in the postprocessing step, specialized adaptive Gaussian and median filters are considered. One in every of the outputs of the binarization step, that shows high recall performance, is used in a proposed postprocessing technique to improve the performance of different binarization methodologies. Finally, we develop a ground truth generation tool, known as PhaseGT, to simplify and speed up the ground truth generation process for ancient document images. The comprehensive experimental results on the DIBCO'09, H-DIBCO'ten, DIBCO'eleven, H-DIBCO'twelve, DIBCO'thirteen, PHIBD'twelve, and BICKLEY DIARY information sets show the robustness of the proposed binarization method on numerous sorts of degradation and document images.