Presentation 2008-02-22
A Study for Caption Character Pattern Extraction
Yutaka KATSUYAMA, Hongliang BAI, Hiroaki TAKEBE, Katsuhito FUJIMOTO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) It is effective to use caption text to search a video from many recorded videos. For caption recognition, it needs to detect caption area and duration, to extract caption character patterns, and to recognize them. This paper reports about caption character patterns extraction. After generating average image of caption duration, color clustering is done to extract decomposed binary images. Then, noise is removed by multi-frame information. The most reliable character line is selected by its OCR result. Some key characters are extracted and combinations of characters are made based on them. The most reliable one is selected by result of character recognition and distance between adjacent characters. We evaluated accuracy using character box coordinates for 83 captions in a news video and got 93.3% recall and 91.6% precision rate.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Caption / Character Recognition / Color Clustering / Noise Removal / Multi Frame
Paper # PRMU2007-239
Date of Issue

Conference Information
Committee PRMU
Conference Date 2008/2/14(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Pattern Recognition and Media Understanding (PRMU)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A Study for Caption Character Pattern Extraction
Sub Title (in English)
Keyword(1) Caption
Keyword(2) Character Recognition
Keyword(3) Color Clustering
Keyword(4) Noise Removal
Keyword(5) Multi Frame
1st Author's Name Yutaka KATSUYAMA
1st Author's Affiliation Software and Solution Laboratories, Language & Media Laboratory, FUJITSU LABORATORIES LTD.()
2nd Author's Name Hongliang BAI
2nd Author's Affiliation Fujitsu Research and Development Center Co., Ltd.
3rd Author's Name Hiroaki TAKEBE
3rd Author's Affiliation Software and Solution Laboratories, Language & Media Laboratory, FUJITSU LABORATORIES LTD.
4th Author's Name Katsuhito FUJIMOTO
4th Author's Affiliation Software and Solution Laboratories, Language & Media Laboratory, FUJITSU LABORATORIES LTD.
Date 2008-02-22
Paper # PRMU2007-239
Volume (vol) vol.107
Number (no) 491
Page pp.pp.-
#Pages 6
Date of Issue