Character Segmentation and Recognition in a Scanned Document
Character Segmentation and Recognition in a Scanned Document
Author(s): Teodor Toshkov, Vladimir Katkalov, Radoslav BorisovSubject(s): ICT Information and Communications Technologies
Published by: Нов български университет
Keywords: OCR; Convolutional neural networks; Projection profile; Gaussian filter; Otsu‘s image segmentation method; Scanline flood fill
Summary/Abstract: The character segmentation and recognition in a scanned document is a part of OCR (Optical Character Recognition), which deals with the problem of digitalization of text in a scanned document. The realization of the software is done by binarizing the image, analyzing the horizontal projection histogram to separate the lines, then separating the characters from each line by analyzing the vertical projection histogram of each line and passing each character to a neural network to be recognized. We are making this research in order to test our ideas for approaching the problem of OCR. There are many documents being digitalized nowadays in different languages and in the field of OCR, there is a lot more potential, which has yet to be researched.
Journal: Computer Science and Education in Computer Science
- Issue Year: 13/2017
- Issue No: 1
- Page Range: 1-19
- Page Count: 19
- Language: English
