Documentation for MDS/VideoText

Technique VideoText
Document MPEG-7 Working Draft MDS
Name Chitra Dorai, Ruud Bolle (IBM T.J. Watson Research Center, New York, USA)
Nevenka Dimitrova, Lalitha Agnihotri (Philips Research, New York, USA)
Email dorai@watson.ibm.com
Contact Chitra Dorai at IBM or Nevenka Dimitrova/Lalitha Agnihotri at Philips
Type Application
External Libraries NONE
Related Ds/DSs VideoText DS
Input MPEG-Video (foo.mpg)
Output foo.xml (in the directory specified with the -b option in the XMWinExe command),
additionally, to any file specified with the -b option, (e.g., to videotext.xml, if specified -btestdir/videotext.xml)
Extraction Yes (IBMVideoTextServer)
Client Appl IBMVideoTextEventBrowserDemoForMPEG-7 Application
Summary The IBM videotext extraction algorithm works by extracting and
analyzing regions in a video frame. The processing stages are
(i) isolating regions that may contain text characters,
(ii) separating each character region from its surroundings and
(iii) verifying the presence of text by consistency analysis across multiple text blocks.
Strong Points The extraction algorithm does very well with low miss
rates, by not missing many of the true text lines in the frame
Limitations NONE
Known Problems Can't handle images larger than HDTV resolution
Parameters The IBMVideoTextServer is coded to handle HDTV frame sizes, 1920x1080.
Changing MAX_SIZE in IBMVideoTextdataTypes.h can customize this
and change related constants found in the same file.