Support for detailed spatio-temporal object queries in a video database system requires the extraction of important objects from the video for indexing; this is impossible to do manually for video databases of realistic size. To address this problem, we propose a method to automatically extract important video objects for object-based indexing. Most of the existing salient object detection approaches detect visually conspicuous structures in images, while our method aims to find regions that may be important for indexing in a video database system. Our method works on a shot basis. We first segment each frame to obtain homogeneous regions in terms of color and texture. Then, we extract a set of regional and interregional color, shape, texture and motion features for all regions. Using these features, we classify each region as being salient or nonsalient using SVMs trained on a few hundreds of example regions. Finally, each salient region is tracked throughout the shot for trajectory generation and consistency check, and final set of salient regions are obtained. Preliminary experimental results are promising. Ongoing work is focused on the extension of the proposed approach for the detection of regions/objects which are semantically more meaningful in terms of human visual perception.

DATE: 31 March, 2008, Monday@ 16:40
PLACE: EA 502