情報処理 , Vol.16 [12], pp.1064-1070 (1975)
Abstract
Effective methods have been proposed for the recognition of the printed
Chinese characters using pattern matching approaches. Through these works, the
positioning errors have been proved to cause fatal deterioration of recognition.
The simplest and most powerful way to rectify the errors may be to seek the
best fit between the input pattern and the standard of each category by
translating one of then in horizontal and/or vertical directions.
We applied the technique to the recognition of printed Chinese characters
whose vocabulary amounts to 1850. Two algorithms belonging to pattern matching
methods were used. On of them is the hierarchical pattern matching method in
frequency domain and the other is the simple binary pattern matching. Applying
the technique of correction, the error rates were reduced remarkably.
The effect of the correction resulted from the remarkable improvement of
intraclass similarities and the slight rises of the interclass similarities.
←[印刷漢字認識]
←[中野の研究紹介]
←[中野の目次]
[情報科学部目次]
| 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