A pictorial dictionary for printed Farsi subwords

Authors:
Afshin Ebrahimi;Ehsanollah Kabir
Affiliations:
Department of Electrical Engineering, Sahand University of Technology (SUT), Tabriz, Iran;Department of Electrical Engineering, Tarbiat Modarres University, Tehran, Iran
Venue:
Pattern Recognition Letters
Year:
2008

Citing 7
Cited 1

A word shape analysis approach to lexicon based word recognition

Pattern Recognition Letters
An Omnifont Open-Vocabulary OCR System for English and Arabic

IEEE Transactions on Pattern Analysis and Machine Intelligence
A new segmentation technique for omnifont Farsi text

Pattern Recognition Letters
A new algorithm for machine printed Arabic character segmentation

Pattern Recognition Letters
Printed PAW Recognition Based on Planar Hidden Markov Models

ICPR '96 Proceedings of the 13th International Conference on Pattern Recognition - Volume 2
Spectral features for Arabic word recognition

ICASSP '00 Proceedings of the Acoustics, Speech, and Signal Processing, 2000. on IEEE International Conference - Volume 06
Experiments with "Characteristic Loci" for Recognition of Handprinted Characters

IEEE Transactions on Computers

Handwritten word spotting in old manuscript images using a pseudo-structural descriptor organized in a hash structure

IbPRIA'11 Proceedings of the 5th Iberian conference on Pattern recognition and image analysis

Quantified Score

Hi-index	0.10

Visualization

Abstract

In this paper, we report on the use of characteristic loci features to cluster printed Farsi subwords, based on their holistic shapes. This yields a pictorial dictionary that can be used in a word recognition system to eliminate the search space. The feature vectors are compressed using PCA. The k-means algorithm is used to cluster 113,340 subwords of 4 fonts and 3 sizes to 300 clusters. The minimum and maximum numbers of cluster members are 59 and 876, respectively. The mean of each cluster is used as its entry in the pictorial dictionary. To evaluate the clustering results, a minimum mean-distance classifier was used to test a set of 5000 subwords. 78.71, 99.01 and 100 percent of these subwords were in the first, first five and first 10 closest clusters, respectively.