Summarization Scheme Based on Near-Duplicate Analysis

David Gorisse 1 Frédéric Precioso 1 Sylvie Philipp-Foliguet 1 Matthieu Cord 2
1 MIDI - Multimedia Indexation and Data Integration
ETIS - Equipes Traitement de l'Information et Systèmes
2 MALIRE - Machine Learning and Information Retrieval
LIP6 - Laboratoire d'Informatique de Paris 6
Abstract : This paper presents our approach to select relevant sequences from raw videos in order to generate summaries to Trecvid 2008 BBC Rush Task. Our system is composed of two major steps: First, the system detects \semantic" shot boundaries and keeps only non-redundant shots; then, the system esti- mates average motion for each shot, as a criterion of amount of information, to better share out the duration of the sum- mary between remaining shots. The rst step is based on a fast near-duplicate retrieval using Locality Sensitive Hashing (LSH) which provides results in few seconds (if we do not take into account decoding and encoding processes). The evaluation of Trecvid shows very promising results, since we ranked 17th over 43 runs, regarding redundancy measure (RE), and 18th for object and event inclusion (IN). These balanced results (most of best teams for the rst criterion are among the latest for the second one) show that our method o ers a quite good trade-o between false negatives (IN) and false positives (RE).
Document type :
Conference papers
Complete list of metadatas

https://hal.archives-ouvertes.fr/hal-00773552
Contributor : Michel Jordan <>
Submitted on : Monday, January 14, 2013 - 11:46:26 AM
Last modification on : Friday, October 4, 2019 - 12:14:02 PM

Links full text

Identifiers

Citation

David Gorisse, Frédéric Precioso, Sylvie Philipp-Foliguet, Matthieu Cord. Summarization Scheme Based on Near-Duplicate Analysis. TVS 2008 - 2nd ACM TRECVid Video Summarization Workshop, Oct 2008, Vancouver, Canada. pp.50-54, ⟨10.1145/1463563.1463571⟩. ⟨hal-00773552⟩

Share

Metrics

Record views

127