Segmentation of Characters of Devanagari Script Documents

Author Manpreet Kaur
Co-Author Ashok Kumar Bathla
DOI

Country : India
Subject : Engineering

Keywords : OCR, Head line, Middle zone, upper zone, lower zone

Abstract

Segmentation of Characters is the final objective of OCR. OCR proceeds through the segmentation procedure by having sequence of steps like input, preprocessing, segmentation. The procedure of segmentation is applied on both handwritten and machine written documents. When the text in any script is single font with single alignment and with single width will be very easy to segment. Because each character has single unique behavior. But the process will be difficult to handle when the text is handwritten. Because while writing using hand various complexities like overlapping, touching or skewness, broken, and non-alignment arises. Due to these reasons the success rate of the segmentation will be reduced. To overcome these issues various segmentation techniques like water reservoir, Collusion dilation algorithms are used. In proposed research segmentation of broken letters of devnagri scripts is taken place. The methodology used is Collusion dilation. This technique has been applied on 500 images dataset of handwritten Devanagari script. The result generated is 96%.

Download