Binarization of Degraded Documents Using Linear Regression

In the age of digitization, it is essential to convert the printed copy of the documents to the digital form. These documents are available through large number of sources - old books, old legal documents, medical-based notes, and safety-related reports. These documents are deteriorated and lead to low contrast. Several approaches have been discussed and designed to overcome the challenges encountered in the binarization of these documents. In this paper, proposed a various challenges encountered in the binarization process and also the method of linear regression in encountered for the removal of various degradation. Common challenges such as Ink bleed, uneven illumination, blurred character and contrast issues have been discussed. The proposed implementation uses linear regression which gives the flexibility to demonstrate the information extraction from the degraded historical documents. In this paper we have used DIBCO data sets from years (2016, 2013, 2012, 2011, 2010 and 2009).