Deformable templates in image sequences

The Bayesian reconstruction of the shape of an object in image sequences using deformable template techniques is considered. The object shape is represented by a prototype outline which is deformed under constraints of continuity and rigidity. The deformation of the template retains consistent shape characteristics whilst allowing variation and evolution of object shape over time. The reconstruction of single images using deformable templates allows incorporation of more detailed prior knowledge of expected image features than a simple local pixel dependence model. An application to a sequence of CCD animated digital images is given.< >