Presentation 1997/12/19
Character Recognition using Automatic Template Selection based on Estimating Degradation
Minako Sawaki, Hiroshi Murase, Norihiro Hagita,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper proposes a method for recognizing low-quality characters in scene images captured by a digital camera and so on. Characters degraded by different illumination conditions and intensity variation are hardly recognized when a single template set obtained from neat learning samples is used as reference patterns. To cope with the problem on the intensity variation, this paper uses several template sets, each of which corresponds to several levels in binarization. A relevant binarization level for input characters is estimated from a table which is trained in advance. Experiments with 441 characters in 12 images show that this method achieves a recognition rate of 82.8%, higher than 76.4% of a single template set.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) character recognition / template selection
Paper # PRMU97-187
Date of Issue

Conference Information
Committee PRMU
Conference Date 1997/12/19(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Vice Chair

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Character Recognition using Automatic Template Selection based on Estimating Degradation
Sub Title (in English)
Keyword(1) character recognition
Keyword(2) template selection
1st Author's Name Minako Sawaki
1st Author's Affiliation NTT Basic Research Laboratories()
2nd Author's Name Hiroshi Murase
2nd Author's Affiliation NTT Basic Research Laboratories
3rd Author's Name Norihiro Hagita
3rd Author's Affiliation NTT Communication Science Laboratories
Date 1997/12/19
Paper # PRMU97-187
Volume (vol) vol.97
Number (no) 459
Page pp.pp.-
#Pages 8
Date of Issue