We present a novel level-set tracking framework that incorporates spatial information.
For tracking, a new rigid registration method is proposed.
For segmentation, a new level-set evolution method is proposed.
Our method shows significantly improved performance in tracking non-rigid objects.