Retrieving event-based semantics from images

This paper discusses the semantic retrieval of image features that capture dynamic aspects or events. Even a single image can demonstrate evidence of dynamic activity. The ability to classify events from one or more images provides a foundation for ultimately extending the capabilities for feature extraction based on real-world happenings. This ability improves a user´s knowledge base and provides insights into a range of possible events that are associated with an image including evidence of unexpected or novel events. In this paper, we describe the requirements for ontologies that support images containing events. Images, as snapshots of dynamic occurrences, require extended ontologies that include event classes and relations. Semantics associated with events give rise to a categorization of image events that can augment image ontology as well as serve as a foundation for event-based querying of images. Categories of image events include: activity-based events, such as blocking and facilitating events; events that are initiating or terminating; events that relate to changes experienced by objects, including appearance, disappearance, and metamorphose events, as well as unexpected events.