Interactive Character Segmentation of Ancient Documents by Controlling Binary Level and its Evaluation
-
- TOMITA Hiroaki
- Ritsumeikan University
-
- SHIBAYAMA Mamoru
- Osaka City University
-
- ARAKI Yoshihiko
- Ritsumeikan University
Bibliographic Information
- Other Title
-
- 古文書画像の2値化レベル制御による対話型文字分割とその評価
- コモンジョ ガゾウ ノ 2チカ レベル セイギョ ニヨル タイワガタ モジ ブ
Search this article
Abstract
Almost all ancient documents were handwritten with cursive styles using writing brush, as result, successive strokes that represent characters are generally connected and individual character patterns are considerably deformed. Character segmentation for documents plays an important role in character recognition and in building image database using such documents. In order to segment each character on a document, we identify the points at which the pressure of the writing brush changes infinitesimally between individual characters. We propose a method for effective segment positioning by automatic binary level control and interactive character segmentation. Result using the above method for character segmentation indicates that the proposed method has an 84.5% ratio of segmentation, which is about 27.2% higher as compared with the result when the method is not used. This paper describes the algorithm for estimating effective segmentation position and result based on this algorithm are evaluated.
Journal
-
- IEEJ Transactions on Electronics, Information and Systems
-
IEEJ Transactions on Electronics, Information and Systems 118 (4), 503-509, 1998
The Institute of Electrical Engineers of Japan
- Tweet
Details 詳細情報について
-
- CRID
- 1390282679584171776
-
- NII Article ID
- 130006843455
- 10002813515
-
- NII Book ID
- AN10065950
-
- ISSN
- 13488155
- 03854221
-
- NDL BIB ID
- 4436279
-
- Data Source
-
- JaLC
- NDL
- Crossref
- CiNii Articles
-
- Abstract License Flag
- Disallowed