1993 Fiscal Year Final Research Report Summary
Construction of a free-formatted document image database system which adds
| Project/Area Number |
03558007
|
|---|
| Research Category |
Grant-in-Aid for Developmental Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Research Field |
Informatics
|
|---|
| Research Institution | KYOTO UNIVERSITY |
|---|
Principal Investigator |
IKEDA Katsuo Kyoto Univ., Information Science, Prof., 工学部, 教授 (30026009)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
西村 真一 三田工業株式会社, 技術部, 開発研究職
OHTA Yuichi Univ.of Tsukuba, Information Sciences and Electronics, Prof., 電子・情報工学系, 教授 (50115804)
HIROSE Shouichi Kyoto Univ., Information Science, Inst., 工学部, 助手 (20228836)
AMANO Akira Kyoto Univ., Information Science, Inst., 工学部, 助手 (60252491)
MINOH Michihiko Kyoto Univ., Integ.Media Env.Exp.Lab., Assoc.Prof., 工学部, 助教授 (70166099)
WATANABE Masako Kyoto Univ., Integ.Media Env.Exp.Lab., Res.Assist. (70127158)
|
|---|
| Project Period (FY) |
1991 – 1993
|
| Keywords | Free Formatted Document / Document Image / Database / Character Recognition / Key Words |
|---|
| Research Abstract |
The purpose of this research is to construct a document image database which can add keywords automatically to free-formatted document image data. Results of this research are as follows :
Color document image processing
We proposed a method for extracting character patterns from color document images assuming that the characters in a sequence has a same color. For character patterns superposed on photographic images, another extraction method was proposed which uses the edges of character patterns. We also presented a method for extracting photographic regions using color information.
Character pattern extraction from free-formatted document images
We proposed a method to make character pattern extraction more precise using character recognition and a word dictionary. A method was also developed for extraction of character pattern sequences with arbitrary slants. We defined a measure of the complexity of characters and related it to the lower bound of the resolution of characters for reliable character recognition.
Document image retrieval methods
We developed a fast and efficient searching algorithm for finding keywords form the storage of character candidates. A method was proposed for helping users finding adequate keywords with the use of automatically extracted words from document images. We also proposed and implemented an interface for retrieving multi media document images using abstract images.
Distributed parallel processing for document image databases
A LAN-based distributed parallel processing environment was constructed for color document image database systems. In this environment, various processing, such as image processing, character recognition, structure analysis, and so on, are considered as modules and parallel execution of modules are implemented.
|
