A region-based method for model-free object tracking

We propose a region-based method for model-free object tracking. In our method the object information of temporal motion and spatial luminance is fully utilized. We first compute the dominant motion of the tracked object. Using this result we warp the object template to generate a prediction template. Static segmentation is incorporated to modify this prediction, where the warping error of each watershed segment and its rate of overlapping with the warped template are utilized to help classification of some possible watershed segments near the object border. Applications to facial expression tracking and two-handed gesture tracking demonstrate the method's performance.