Proc. IJCPR, pp.119-121 (1978)
Abstract
A new optical Chinese character reader, Kanji OCR. is presented. New computer
application fields require Japanese text reading devices which promote office
automation or computerization of large administrative systems. The present Kanji
OCR aims at reading the Chinese characters, in up 2,400 categories, as printed
by a Japanese typewriters.
A recognition scheme of hierarchical pattern matching is adopted to gain high
recognition performance and hardware simplicity. A new algorithm of
hill-climbing matching is introduced to reduce positioning errors. A contextual
preprocessing technique is also proposed that reduces the number of candidates
before classification. The contextual preprocessing speeds up recognition time
and improve the accuracy. Experiments on in-house and field samples have yielded
the error rate of the order 10-4 and rejection of 10-3.
| mail address: | |
← お手数ですが打ち込んで下さい |
First Written Before June 17, 1998
Transplanted to KSU Before May 16, 2003
Transplanted to So-net May 3, 2005
Last Update April 8, 2007
© Yasuaki Nakano 1998-2007