A Modified Wellner’s based Binarization on Ancient Malayalam Documents

Abstract

Preservation of precious knowledge present in Malayalam literature is a tedious task because of the presence of a huge amount of degradations in historical documents. One way to save these documents is to enhance the Malayalam manuscripts and storing them electronically. Here we are using historical Malayalam documents like poems, agreement copies and palm leaves as experimental dataset. In our proposed work is a novel binarization model which is based on modified Wellner’s algorithm. The degraded input image is converted into an integral image and then a modified version of Wellner’s algorithm is applied to it. This would enable us to enhance the document and further proceed with remaining phases of OCR. In binarization mainly we are focusing on the challenges from degraded documents like Non-uniform Background Illumination, Stains, bleed through etc. In our work we are developing a model which effectively binarizes the degraded Malayalam documents, especially the challenges like oil stain, smudge, and uneven illuminations. Finally, we will be able to validate and check the accuracy of the proposed model effectively. The proposed algorithm yielded an accuracy of 92%.