Ben Hutchings wrote:
> Jani PATOKALLIO wrote:
> > The flip side of the coin is that Japanese OCR has to deal with
> > 2000+ characters as opposed to 52, and rapid Japanese handwriting
> > tends to degenerate into a scrawl that even the natives have problems
> > dealing with.
> <snip>
>
> Yet there is apparently a high-speed OCR system in use for sorting
> mail in Japan. There's a very brief description of this at
> http://www.toshiba.co.jp/tech/review/1998/high98/information/i8/
The contents would seem to hint that the OCR is applied to the
7-digit postal codes, not the addresses themselves.
Recognizing handwriting with fixed-size characters, known stroke
direction/order and a given input device (as in the Zaurus) is already
tough enough a task; when you've got wildly variant sizes, styles,
orientations, colors and thicknesses, just managing to read numbers
is an impressive achievement...
Cheers,
--
Jani PATOKALLIO / jpatokal@iki.fi / +81 90 7722 3557
Sanpo Laboratory, Mechano-Informatics Dept., University of Tokyo
ヤニ・パトカリオ / jani@sanpo.t.u-tokyo.ac.jp / 090 7722 3557
東京大学、工学系研究科、機械情報工学科、算法設計研究室
[ Did you check the archives? http://www.appelsiini.net/keitai-l/ ]
Received on Fri Jan 26 06:02:26 2001