CVSL Logo
FrancaisHome
AboutPeopleResearchPublicationsEventsProfile
About
Publications

 

 

 

CERVIM

REPARTI

MIVIM

Allograph Adjacency Constraints for Cursive Script Recognition


Marc Parizeau and Réjean Plamondon


Abstract - This paper de nes horizontal and vertical adjacency con- straints between allographs. The constraints are used to build an allograph segmentation graph containing paths that represent character strings possibly corresponding to an unknown cursive word. Allographs are modelled by fuzzy-shape grammars and are segmented within this cur- sive word by an adequate parser. Results are given for multi-writer random letter sequences. For a test database containing cursive handwriting samples from 10 different writers, and without using any linguistic knowledge, average character recognition rates of 84.4% to 91.6% are obtained depending on whether only the first string outputted by the system is considered, or if the best of the top ten is accepted, the recognition system being non per- sonalized.

download document

Bibtex:

@inproceedings{Parizeau64,
    author    = { Marc Parizeau and Réjean Plamondon },
    title     = { Allograph Adjacency Constraints for Cursive Script Recognition },
    booktitle = { Proc. of the Third International Workshop on Frontier in Handwriting Recognition (IWFHR) },
    pages     = { 252-261 },
    year      = { 1993 },
    month     = { May 25-27 },
    location  = { Buffalo (USA) }
}

Last modification: 2002/06/14 by parizeau

     
   
   

©2002-. Computer Vision and Systems Laboratory. All rights reserved