armin@ai.toronto.edu (Armin Haken) (03/10/88)
There will be an AI Seminar in room GB244 at 11AM next Wednesday, March 16, by Demetri Terzopoulos. Please note the unusual time, and the fact that the room has been changed to GB244. The speaker is a candidate for a position here. U of T is competing with various other institutions for Dr. Terzopoulos, and we want to make the best possible impression. Part of that impression depends on good attendance at the talk. DEFORMABLE MODELS IN COMPUTER VISION AND GRAPHICS Demetri Terzopoulos Schlumberger Palo Alto Research Vision and graphics are mutually converse disciplines; the former is concerned with the analysis of images, the latter with their synthesis. They pose similar problems at the object modeling level. I shall describe a physically-based approach to analyzing and synthesizing the shapes and motions of nonrigid objects. Objects are modeled using deformable curve, surface, and solid primitives, while constraints are represented as dynamic forces applied to these primitives. In the context of graphics (the direct problem), realistic images of flexible objects may be synthesized when the applied forces arise from the interaction of deformable models with simulated physical environments. With regard to vision (the inverse problem), deformable models may be used to infer the shapes and motions of objects from their images. Here the forces are derived from natural image data and enforce image-based constraints. They actively shape and move models to achieve maximal consistency with imaged objects of interest and to maintain the consistency over time. I will present results of applying deformable models to image contour extraction, stereo and motion correspondence matching, static 3D object reconstruction from monocular images, and the recovery of 3D shape and nonrigid motion of objects from dynamic stereo imagery. The video presentation will include computer animations of deformable models reacting to a variety of simulated physical phenomena. -- || Armin Haken armin@ai.toronto.edu || || UofT DCS, Toronto, M5S-1A4 (416)978-6277 ||