Seeded Semantic Object Generation toward Content-Based Video Indexing

A novel semantic object generation algorithm is proposed by collaborative integration of a new feature-based image segmentation technique and a seeded region aggregation procedure. The homogeneous image regions with closed boundaries are obtained by integrating the results of color edge detection and seeded region growing procedures. The object seeds are then distinguished from these homogeneous image regions. The semantic objects are finally generated by a seeded region aggregation procedure. The extracted semantic objects can then be tracked along the time axis. The video objects can then be used as the basic units for content-based video indexing.