Appraisal of localized binarization methods on Tamil palm-leaf manuscripts

Palm-leaf manuscripts have enormous information, very much essential to day-to-day life. Due to climatic changes and aging factors these manuscripts have started degrading. To preserve the information present on the palm-leaves, digital images of each leaves have been taken up and stored. These images can be viewed as and when required. However, the storage volume required for each leaf is tremendously large, thereby increasing the cost. In order to have only the textual information for storage and segmentation purposes, it is mandatory that these images are binarized. Binarization of degraded document images has been a challenge to many computer scientist. This paper primarily deals with binarization of the palm-leaf manuscript images using Niblack, Sauvola, NICK and Bradley binarization algorithms. Binarization not only reduces the size, thereby protecting the textual content, but also helps in next level of character recognition which includes the segmentation.