Presentation 2007-09-28
Subtitle Detection from Video Data : Soft Decision for Subtitle Candidates Based on a Markov Model
Takaaki KASHIO, Yutaka NAKANO, Toshiyuki YOSHIDA,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) This paper proposes a new technique for subtitle region detection which operates without full-decoding for MPEG bitstreams. As the feature values for the detection, we focus on a variation of the reference frame for B-picture macroblocks (MBs) and the DCT coefficients for I-picture DCT block. These values are fed into a Markov Model(MM) as a temporal sequence to detect the subtitle regions from its state transition. The experimental results given in this paper show that our technique gives about 74.7% and 94.3% of the recall and the precision ratios, respectively, for some test sequences.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) MPEG bitstream / subtitle region detection / Markov-model
Paper # SIP2007-103,SIS2007-39,SP2007-65
Date of Issue

Conference Information
Committee SIS
Conference Date 2007/9/21(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 Smart Info-Media Systems (SIS)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Subtitle Detection from Video Data : Soft Decision for Subtitle Candidates Based on a Markov Model
Sub Title (in English)
Keyword(1) MPEG bitstream
Keyword(2) subtitle region detection
Keyword(3) Markov-model
1st Author's Name Takaaki KASHIO
1st Author's Affiliation Dept. Information Science, Faculty of Eng., University of Fukui()
2nd Author's Name Yutaka NAKANO
2nd Author's Affiliation Dept. Information Science, Faculty of Eng., University of Fukui
3rd Author's Name Toshiyuki YOSHIDA
3rd Author's Affiliation Dept. Information Science, Faculty of Eng., University of Fukui
Date 2007-09-28
Paper # SIP2007-103,SIS2007-39,SP2007-65
Volume (vol) vol.107
Number (no) 237
Page pp.pp.-
#Pages 4
Date of Issue