Statistical modeling for motion-based video classification and retrieval
Résumé
We have developed an original approach for content-based video indexing and retrieval. By introducing a causal Gibbsian modeling of the spatio-temporal distribution of appropriate local motion-related measurements, we have designed a general and efficient statistical framework for non parametric motion modeling, motion recognition and classification, and motion segmentation. It is exploited for motion-based video indexing and video retrieval for both global and partial queries by example.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...