A critical problem in automatic segmentation of natural video is accurate identification of object boundaries. We present an algorithm of reasonable complexity, where the primary novelties consist in a method for motion-based object tracking and a method for object boundary identification employing edge-oriented morphological processing, both at relatively low complexity. Experimental results show that the algorithm achieves good performance except for videos with violently moving backgrounds and low contrast between the moving object and the background. The required computing time of the algorithm is seen to be appropriate for real-time desktop or portable multimedia applications.
|Number of pages||8|
|Journal||International Journal of Electrical Engineering|
|State||Published - 1 Aug 2005|
- Alpha plane
- Object extraction
- Video segmentation