Probabilistic models of shape

The purpose of this second part of the book is to put Active Contours into a probabilistic setting. As chapter 2 claimed, the probabilistic framework is essential for dealing with classes of shapes and motions. It is valuable even with deformable templates, in static problems, to describe classes of shapes. Then probabilistic modelling is extended to dynamic problems, to mesh with the powerful Kaiman filtering formalism, in which cumulative temporal uncertainty about shape is counterbalanced by the inflow of measurements from an image sequence.