Please use this identifier to cite or link to this item:
Title: An integrated method for classification of indus and english document images
Authors: Kavitha, A.S.
Shivakumara, P. 
Hemantha Kumar, G.
Keywords: Cursiveness
English document classification
Indus document classification
Region growing
Issue Date: 2014
Citation: Kavitha, A.S.,Shivakumara, P.,Hemantha Kumar, G. (2014). An integrated method for classification of indus and english document images. Lecture Notes in Electrical Engineering 248 LNEE : 343-355. ScholarBank@NUS Repository.
Abstract: Unlike classification of documents with plain background and high resolution, classification of historical document, namely Indus script written on stone, wall, and palm leaves is challenging because of sources on which script is written and various handwriting, which causes noise, distortions, background variations, multisized text, and multifont. In this paper, we propose an integrated method that has two-stage algorithms to classify Indus and English from the South Indian documents. The first stage uses morphological operations and thinning on Canny of the input image to study the straightness and cursiveness of thinned components to classify the Indus document from the South Indian and English. The second stage proposes region growing and thinning to study the straightness and cursiveness of the thinned edges to classify the English from the South Indian documents. We select 100 documents for each script in total 600 documents to evaluate the performance of the method. The comparative study with existing method shows that the proposed method outperforms the existing method in terms of classification rate. © 2014 Springer India.
Source Title: Lecture Notes in Electrical Engineering
ISBN: 9788132211563
ISSN: 18761119
DOI: 10.1007/978-81-322-1157-0_35
Appears in Collections:Staff Publications

Show full item record
Files in This Item:
There are no files associated with this item.


checked on Nov 15, 2018

Page view(s)

checked on Nov 9, 2018

Google ScholarTM



Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.