The goal of this dissertation is to present a method for motion reconstruction. Motion carries a lot of information about spatio-temporal relationships between image objects. Its reconstruction plays a very important role in many research areas.
The approach is based on image moments and the bsptree data structure. In each frame we construct a bsptree to store a set of ellipses that approximates the object’s shape. The hierarchical structure, as well as the set of ellipses, are computed using invariant properties of image moments. These ellipsoidal structures are matched, which allows to track the object. Motion is represented by geometric transformations between the sets of ellipses and stored in the hierarchical structure. Motion reconstruction is done by interpolating these transformations.
Following, we propose an application for our method. The approach is used to produce intermediate object’s shape in an image sequence.