The retrieval of videos of interest from large video collections is a main open problem which calls for the definition of new video content characterization techniques in term of both visual descriptors and semantic annotations. In this paper, we present an efficient and effective video retrieval system which profitably exploits the functionalities offered by a semantic-based automatic video annotator using video shots similarity to suggest relevant labels for the videos to be annotated. Similarity queries based on semantic labels and/or visual features are implemented and experimentally compared on real data in order to measure the retrieval contribution of each type of video content information.
Efficient and Effective Similarity-based Video Retrieval
BARTOLINI, ILARIA;
2010
Abstract
The retrieval of videos of interest from large video collections is a main open problem which calls for the definition of new video content characterization techniques in term of both visual descriptors and semantic annotations. In this paper, we present an efficient and effective video retrieval system which profitably exploits the functionalities offered by a semantic-based automatic video annotator using video shots similarity to suggest relevant labels for the videos to be annotated. Similarity queries based on semantic labels and/or visual features are implemented and experimentally compared on real data in order to measure the retrieval contribution of each type of video content information.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
