Options
1999
Conference Paper
Titel
A New Approach for Geometrically Deformable Models by Integration of Knowledge
Abstract
Deformable models often require some degree of user interaction to produce an accurate reconstruction of an object. Due to noise and artifacts which are inherent in most data sets, automation is a very problematic task. Case specific a priori information may increase the chances of a successful solution by effectively reducing the solution space of the problem to those solutions which share similar characteristics with that information. In this paper, we present a new approach for the geometrically deformable models by integration of a priori information. Besides local and global geometric properties this comprises shape knowledge as well as intuitive knowledge and structural features. We propose its modelling in an energy minimising concept by free function assignment. The power of this method is illustrated using automatic segmentation of synthetic and medical images, reconstruction of destroyed teeth from range images, and labelling of anatomical structures.
Konferenz