Skew correction of document images using interline cross-correlation
CVGIP: Graphical Models and Image Processing
Character Recognition of Arabic and Latin Scripts
IV '00 Proceedings of the International Conference on Information Visualisation
A Reconfiguarable Printed Character Recognition System Using a Logic Synthesis Tool
EUROMICRO '98 Proceedings of the 24th Conference on EUROMICRO - Volume 1
A Novel Approach for Skew Estimation of Document Images in OCR System
CGIV '05 Proceedings of the International Conference on Computer Graphics, Imaging and Visualization
Robust Skew Detection in mixed Text/Graphics Documents
ICDAR '05 Proceedings of the Eighth International Conference on Document Analysis and Recognition
Robust detection of skew in document images
IEEE Transactions on Image Processing
Fourth-order partial differential equations for noise removal
IEEE Transactions on Image Processing
Hi-index | 0.00 |
Template matching is vital for any pattern recognition application but its computational cost increases as templates pool increases. We propose a novel two stage algorithm for the recognition of printed segmented isolated characters of any script especially for cursive scripts like Arabic and Urdu. This algorithm is based on cluster guided pattern recognition through the use of k-means clustering algorithm to considerably reduce the recognition time for the template matching based pattern recognizer. In the first stage Confusion matrix of templates pool is generated through template matching and then applying cluster prediction algorithm on confusion matrix, three different sets of clusters are predicted using k-means clustering. Based on a cost evaluation criterion, the final value for clusters is selected which is used to perform clustering. The proposed approach automatically ensures the same recognition rate as that of template matching alone because cluster centers just serve to reduce the search space of template matching from global pool of templates to local one and actual recognition is still function of template matching. The proposed technique is applied on Urdu character set in Naskh font style. The implementation results show 8.45 times faster recognition rate.