Detecting and Filling Double Text in Korean Signboard Images

Author

Dinh, Toan Nguyen ; Huy Le Phat ; Lee, Gueesang

Author_Institution

Dept. of Electron. & Comput. Eng., Chonnam Nat. Univ., Gwangju, South Korea

Volume

1

fYear

2009

fDate

11-14 Oct. 2009

Firstpage

139

Lastpage

144

Abstract

Double boundary text, also called outlined text, problem is presented when the text boundary has different color with background color and text stroke color. This problem, which frequently occurs in natural text of signboards, imposes another obstacle to the proper extraction of text information. In this paper, an efficient method is proposed based on well-known filling algorithms and characteristics of double boundary text for detecting and correcting double boundary text problem. The experimental results attained in different natural signboard images show the efficiency of our method.

Keywords

character recognition; feature extraction; image colour analysis; text analysis; Korean signboard images; background color; double boundary text; double text detection; filling algorithms; outlined text; text information extraction; text stroke color; Computational complexity; Data mining; Filling; Frequency domain analysis; Gray-scale; Image color analysis; Image segmentation; Information technology; Layout; Neural networks;

fLanguage

English

Publisher

ieee

Conference_Titel

Computer and Information Technology, 2009. CIT '09. Ninth IEEE International Conference on

Conference_Location

Xiamen

Print_ISBN

978-0-7695-3836-5

Type

conf

DOI

10.1109/CIT.2009.14

Filename

5329390