Bangla text extraction by digital image processing

Abstract

Optical character recognition (OCR) is a technology to extract the text from an image. The main purpose of an OCR is to make editable text documents from different scanned documents, image files or books. In this paper, we would discuss the process to develop an OCR for Bangla language. Bangla script contains different shapes and sizes of text. Therefore, extraction of Bengali text from images becomes challenging. In this paper, we would discuss the process of developing an OCR for Bengali language, we focus on the training data preparation process, Tesseract integration procedure for character recognition and the post-processing techniques. Before the recognition step, few preprocessing steps are needed like noise removal, convert to gray scale and binarization for scanned documents. In this paper, we present the basic steps required for developing a Bangla OCR and a complete workflow for development process with the probable errors encountered during recognition using several techniques. We used Tesseract version 3.04 for windows operating system and ‘NIKOSH’ Bangla font in this project. For clear documents, around 95% word level recognition accuracy has been obtained.