| Conference/Proceedings | Proceedings of the IEEE International Conference on Multimedia & Expo (ICME) |
| Start date | 11.07.2011 |
| End date | 15.07.2011 |
| Address | Barcelona, Spain |
| Author(s) | Marina Georgia Arvanitidou, Michael Tok, Andreas Krutz and Thomas Sikora |
| Title | Short-Term Motion-Based Object Segmentation |
| Abstract | In this paper we present an automatic motion-based object segmentation algorithm for video sequences with moving camera. For every frame, two error frames are generated using short-term motion compensation, they are combined, and a thresholding segmentation algorithm is applied. Recent advances in global motion estimation eliminate outliers in the background area, and thus enable more precise definition of the foreground. We show that employing only the motion information of two adjacent frames we achieve improved results compared with a previously proposed short-term motion-based method and we provide subjective and objective evaluation. |
| Key words | multimedia analysis, object segmentation, camera motion estimation, thresholding |
| File | 1290Arvanitidou2011.pdf |