In order to improve scene understanding, the focus in image and video retrieval research has shifted from low-level feature extraction and analysis to producing high-level semantic representation of scenes. A framework for modeling semantic context in images is proposed to facilitate scene classification and at the same time to detect novel scenes. Besides novelty detection in image collections, the proposed semantic context can also be useful for key-frame extraction in video sequences. We will present some experiments conducted on wildlife scenes and video clips.
Last modified: Wednesday, 22-Sep-2010 15:05:23 NZST
This page is maintained by the seminar list administrator.