A Novel Trajectory Clustering Approach for Motion Segmentation

We propose a novel clustering scheme for spatio-temporalsegmentation of sparse motion fields obtained from feature tracking. Theapproach allows for the segmentation of meaningful motion componentsin a scene, such as short- and long-term motion of single objects, groupsof objects and camera motion. The method has been developed within aproject on the analysis of low-quality archive films. We qualitatively andquantitatively evaluate the performance and the robustness of the ap-proach. Results show, that our method successfully segments the motioncomponents even in particularly noisy sequences.