Feature Extraction by Hierarchical Overlapped Elastic Meshing for Handwritten Chinese Character Recognition

Authors:
Tianlei Wu;Shaoping Ma
Affiliations:
-;-
Venue:
ICDAR '03 Proceedings of the Seventh International Conference on Document Analysis and Recognition - Volume 1
Year:
2003

Citing 4
Cited 2

A nonlinear normalization method for handprinted Kanji character recognition—line density equalization

Pattern Recognition
A Handwritten Character Recognition System Using Directional Element Feature and Asymmetric Mahalanobis Distance

IEEE Transactions on Pattern Analysis and Machine Intelligence
OCR in a Hierarchical Feature Space

IEEE Transactions on Pattern Analysis and Machine Intelligence
High Accuracy Handwritten Chinese Character Recognition by Improved Feature Matching Method

ICDAR '97 Proceedings of the 4th International Conference on Document Analysis and Recognition

Multi-lingual offline handwriting recognition using hidden Markov models: a script-independent approach

SACH'06 Proceedings of the 2006 conference on Arabic and Chinese handwriting recognition
Zoning methods for handwritten character recognition: A survey

Pattern Recognition

Quantified Score

Hi-index	0.00

Visualization

Abstract

Meshing is a popular technology for feature extractionin handwritten Chinese character recognition. In thispaper, we propose a hierarchical overlapped elasticmeshing (HOEM), which integrates overlap, fuzzymembership and hierarchical division into elasticmeshing. The overlapping zones and the fuzzymemberships of the pixels in each zone are dynamicallycalculated through mapping the input image to a virtualnormalized image. Such mapping is fulfilled by nonlinearshape normalization based on weighted dot density.Furthermore, using hierarchical division, the globaloverlapped elastic meshing (GOEM) is extended to theproposed HOEM. Experiments are conducted on theETL9B database to compare different meshing methods.The results indicate that the proposed method achieves ahigher recognition rate than conventional meshingmethods.