A Hybrid Ostu based Niblack Binarization for Degraded Image Documents

Many Historical Documents have been preserved by libraries to protect these documents and to recover knowledge. These old documents suffer from different types of degradations, some of them are uneven illuminations, bleeding through the blur, ink stains, and different sorts of degradation. Thresholding technique is a significant phase in the image processing stage for document image scanning and processing. The main objective of binarization is to extract characters and eliminate noise from the original images. The proposed method is a classifying system to consolidate various binarization techniques for degraded document images by modifying the local and global binarization. The proposed algorithm consists of two phases: modified Otsu's and Niblack's techniques, and consolidates the advantages of modified local and global binarization techniques. The datasets used for proposed work are DIBCO 2009-2019, ICDAR 2019, and self-gathered situated image sets. Compared with two other classical algorithms, using three different evaluation techniques such as F-measure, PSNR and NRM.