Mining TV Broadcasts for Recurring Video Sequences

  • We introduce an algorithm and a real-time system for mining TV broadcasts for recurring video sequences. The algorithm is frame-accurate, i.e., it exactly identifies with which frame a repeating sequence starts and ends resulting in a temporal accuracy of 40ms for PAL videos and 33ms for NTSC videos. The algorithm is also efficient. A 24-hour live-stream can be processed on a standard PC in less than 4 hours including the computational expensive video decoding. This efficiency is partially achieved by means of an inverted index for identifying similar frames rapidly. Images are mapped to the index by first calculating a gradient-based image feature, which in turn is mapped to the index via a hash function. The search algorithm consists of two steps: (1) searching for recurring short segments of e.g. 1 second length (called clips), and (2) assembling these small segments into sets of repeating long and complete video sequences. In our experiments we investigate the sensitivity of theWe introduce an algorithm and a real-time system for mining TV broadcasts for recurring video sequences. The algorithm is frame-accurate, i.e., it exactly identifies with which frame a repeating sequence starts and ends resulting in a temporal accuracy of 40ms for PAL videos and 33ms for NTSC videos. The algorithm is also efficient. A 24-hour live-stream can be processed on a standard PC in less than 4 hours including the computational expensive video decoding. This efficiency is partially achieved by means of an inverted index for identifying similar frames rapidly. Images are mapped to the index by first calculating a gradient-based image feature, which in turn is mapped to the index via a hash function. The search algorithm consists of two steps: (1) searching for recurring short segments of e.g. 1 second length (called clips), and (2) assembling these small segments into sets of repeating long and complete video sequences. In our experiments we investigate the sensitivity of the algorithm concerning all system parameters and apply it to the detection of unknown commercials within 24 and 48 hours of various TV channels. It is shown that the method is an excellent technique for searching for unknown commercials. Currently, the system is used 24 hours 7 days a week in various countries to log all broadcast commercials fully automatically.show moreshow less

Download full text files

Export metadata

Statistics

Number of document requests

Additional Services

Share in Twitter Search Google Scholar
Metadaten
Author:Ina Döhring, Rainer LienhartGND
URN:urn:nbn:de:bvb:384-opus4-11128
Frontdoor URLhttps://opus.bibliothek.uni-augsburg.de/opus4/1323
Series (Serial Number):Reports / Technische Berichte der Fakultät für Angewandte Informatik der Universität Augsburg (2009-03)
Type:Report
Language:English
Publishing Institution:Universität Augsburg
Release Date:2009/10/29
Tag:frame-accurate video mining; recurring video clips; commercial detection
Institutes:Fakultät für Angewandte Informatik
Fakultät für Angewandte Informatik / Institut für Informatik
Fakultät für Angewandte Informatik / Institut für Informatik / Lehrstuhl für Maschinelles Lernen und Maschinelles Sehen
Dewey Decimal Classification:0 Informatik, Informationswissenschaft, allgemeine Werke / 00 Informatik, Wissen, Systeme / 004 Datenverarbeitung; Informatik
Licence (German):Deutsches Urheberrecht