Document Image Binarization Process

Marcel Prodan, Costin-Anton Boiangiu

Abstract


Technology has made significant strides in recent years, which accounts for how pervasive it is in our daily lives. In order to address the fundamental issue with historical document preservation, namely their degeneration, this work suggests using new technology. The method is built on pieces of artificial intelligence that can read the writing from a page and recognize the useful information, converting it into a digital version. Contrary to photographing or scanning, binarizing a document is a considerably more effective method, both in terms of quality—the legibility of the writing—and quantity—the amount of memory needed to retain the resulting image. According to common assessment measures, the suggested fully convolutional network manages to deliver results that are comparable to those of other solutions of a similar nature.

Keywords


image binarization, artificial intelligence, convolutional neural network, deep neural network, fully convolutional network.

Full Text:

PDF


(C) 2010-2022 EduSoft