Restoration of characters in degraded inscriptions using phase based binarization and geodesic morphology

Abstract

It is the requirement of the time to store and conserve the ancient manuscripts for the use of next generation. Epigraphists find it hard to decrypt the information present in inscriptions due to variety of reasons including erosion of letters, noise and many more. Here, we present a new binarization and postprocessing technique to efficiently extract and reconstruct the foreground text from heavily degraded documents. The proposed method uses combination of phase based feature maps and geodesic morphology with anisotropic filtering. Phase feature maps will binarize the text by removing background noise and geodesic operators will reconstruct the deteriorated characters. Statistical performance evaluation is done on different datasets and efficiency of the proposed method is demonstrated by comparing with many state of art algorithms.