DocumentCode :
627277
Title :
Computer vision vs human perception: Novel preprocessing technique to reduce inter-character similaity of Bangla alphabet
Author :
Chowdhury, T.M. ; Naser, M.A. ; Ahmed, Foisal
Author_Institution :
Dept. of Electr. & Electron. Eng., Islamic Univ. of Technol. (IUT), Gazipur, Bangladesh
fYear :
2013
fDate :
17-18 May 2013
Firstpage :
1
Lastpage :
5
Abstract :
In Bangla alphabet, most of the characters share same features and such similarity misleads a recognizer as it makes decision based on measures of absolute difference. A preprocessing step called `scrambling´ and modification of existing grouping scheme are proposed to decrease the intercharacter similarity among Bangla characters. The theory behind this proposal is originally inspired by experiments done on human vision and perception by psychologists and vision scientists. Experimental results corroborate theoretical predictions which lead us to propose scrambling and new grouping technique as effective steps in character recognition.
Keywords :
character recognition; computer vision; vision; Bangla alphabet intercharacter similarity reduction; absolute difference measure; character recognition; computer vision; grouping scheme modification; grouping technique; human perception; human vision; preprocessing technique; scrambling step; Accuracy; Character recognition; Computer vision; Computers; Feature extraction; Shape; Support vector machines; Bangla character recognition; computer vision; feature extraction; human perception; preprocessing;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Informatics, Electronics & Vision (ICIEV), 2013 International Conference on
Conference_Location :
Dhaka
Print_ISBN :
978-1-4799-0397-9
Type :
conf
DOI :
10.1109/ICIEV.2013.6572630
Filename :
6572630
Link To Document :
بازگشت