Proceedings of 1st ICPR, pp.172-178 (1973)
Summary
The authors have already reported an algorithm for the optical recognition of printed Chinese and Japanese kana characters of single font. The recognition scheme is based on the correlation between the unknown input and the standard patterns in the main frequency components of amplitude spectra of two projection profiles on x-y orthogonal axes. The only defect of the algorithm is the deterioration of recognition when the stroke-width varies.
The first part of this paper describes how to rectify the defect. The improvements proposed for this purpose are based on the use of an optimum frequency range, compensation of the stroke-width variation on the frequency domain and the increase of projection directions.
In the second part of this paper, the authors propose an algorithm of quasi-two-dimensional elastic matching for the recognition of multi-font printed Chinese characters.
| mail address: | |
← お手数ですが打ち込んで下さい |
First Written Before June 16, 1998
Transplanted to KSU Before May 15, 2003
Transplanted to So-net May 3, 2005
Last Update April 8, 2007
© Yasuaki Nakano 1998-2007