A Generative Model For Dynamic Canvas Motion

Matthew Kaplan and Elaine Cohen, University of Utah

Abstract

We present techniques for constructing realistic canvas and paper models and for enabling interactive dynamic canvas motion. Dynamic canvas motion means that there is a correspondence between the motion of canvas features and the motion of the models in the scene. Our artificial paper is created by simulating the physical process of creating paper with many individual fibers. To enable canvas motion, fibers are associated with each of the models in the scene. At runtime, the fibers associated with visible portions of the models and background fibers are used to construct a 2D canvas. Because fibers are tied to the models, the motion of canvas feature corresponds to the motion of each model. This allows us to match the motion field of our dynamic 2D canvas to that of the the 3D scene exactly.

Computational Aesthetics 2005
The full paper

Sample images (click for a larger view):

Some animations (fair warning : I had to compress these like crazy to get them small enough to fit on the server, so you'll see some pixel flicker errors that aren't present in the originals)

Bunny - ID method
Bunny - Stencil method
Menagerie - Rotation with ID method
Menagerie - Rotation with ID method 2nd pass
Menagerie - Rotation with stencil method
Menagerie - Translation with ID method
Menagerie - Translation with ID method 2nd pass
Menagerie - Translation with stencil method