Generation and segmentation of motion trajectories

Presents an algorithm for selecting and linking interesting flow vectors across a sequence of frames for computing motion trajectories. The authors track tokens which have both interesting pixel gray values in the spatial domain and in the optical flow field in the temporal domain. This and operation effectively removes some redundant trajectories. Due to errors introduced during the computation of optical flow, and the linking of such flow vectors across a sequence of frames, the resultant trajectories are not always smooth. The authors discuss a Kalman filtering based approach for smoothing the trajectories. This approach is also able to deal with the occlusion of feature points. The authors also present a simple algorithm for segmenting motion trajectories