Data-based motion simulation through a Green's function approach

We describe a data-driven motion simulation algorithm based on an affine image-matching equation. Solving such equation via the Green's function approach, we have obtained a pair of filters which, when applied over an input image, allow the generation of virtual sequences that convey a compelling motion impression. Complex rigid and nonrigid motions have been simulated this way, such as those of a pulsating heart, a twig in the wind and a levitating ball.