Structure and motion estimation from dynamic silhouettes under perspectiveprojection

We address the problem of estimating the structure and motion of a smooth c urved object from its silhouettes observed over time by a trinocular stereo rig under perspective projection. We first construct a model for the local structure along the silhouette for each frame in the temporal sequence. Th e local models are then integrated into a global surface description by est imating the motion between successive time instants. The algorithm tracks c ertain surface features (parabolic points) and image features (silhouette i nflections and frontier points) which are used to bootstrap the motion esti mation process. The entire silhouettes along with the reconstructed local s tructure are then used to refine the initial motion estimate. We have imple mented the proposed approach and report results on real images.