Generation of templates for low-resolution text recognition using a hypothesis graph
- 16 December 2008
- journal article
- Published by Pleiades Publishing Ltd in Pattern Recognition and Image Analysis
- Vol. 18 (4) , 638-642
Abstract
We propose a recognition method of character-string images captured by portable digital cameras. A challenging task in character-string recognition is the segmentation of characters. In the proposed method, a hypothesis graph is used for recognition-based segmentation of the character-string images. The hypothesis graph is constructed by the subspace method, using eigenvectors as conditionally elastic templates. To obtain these templates, a generation-based approach is introduced in the training stage. Various templates are generated to cope with low-resolution. We have experimentally proved that the proposed scheme achieves high recognition performance even for low-resolution character-string images.Keywords
This publication has 4 references indexed in Scilit:
- Recognition of Screen-Rendered Text2006
- Camera-based analysis of text and documents: a surveyInternational Journal on Document Analysis and Recognition (IJDAR), 2005
- An Evaluation of Statistical Approaches to Text CategorizationInformation Retrieval Journal, 1999
- A survey of methods and strategies in character segmentationIEEE Transactions on Pattern Analysis and Machine Intelligence, 1996