Integrated morphological image information method for complex transformed images

Author

Takahashi, Kunio ; Amanuma, Hiroshi ; Miyazaki, Takayuki ; Chiba, Shuuhou

Author_Institution

Fac. of Eng., Kanagawa Univ., Yokohama, Japan

Volume

2

fYear

2000

fDate

2000

Firstpage

983

Abstract

This paper presents new recognition method for handwritten Chinese characters. A number of integrated morphological image information based on complex transformed images and subareas is used to get a good recognition rate. Similarity by morphological features are used for recognition. Complex images are obtained by the distance transform and nonlinear normalization. Instead of selective use of complex image information and integrated image information, the general similarity of each similarity based on complex images, subareas and integrated image information is used to get good recognition. General similarity is based on the deviation value of each similarity. We obtained a recognition rate of 99.71% for ETL-9, the best in the world

Keywords

character recognition; mathematical morphology; transforms; character recognition method; complex transformed images; distance transform; general similarity; handwritten Chinese characters; integrated morphological image information; morphological features similarity; nonlinear normalization; recognition rate; similarity deviation value; Character recognition; Dictionaries; Feature extraction; Handwriting recognition; Image databases; Image recognition; Spatial databases;

fLanguage

English

Publisher

ieee

Conference_Titel

Signal Processing Proceedings, 2000. WCCC-ICSP 2000. 5th International Conference on

Conference_Location

Beijing

Print_ISBN

0-7803-5747-7

Type

conf

DOI

10.1109/ICOSP.2000.891689

Filename

891689