Physics-based flow estimation of fluids

We present a physics-based method to compute the optical flow of a fluid. In most situations, gray level changes in an image do not provide sufficient information to completely ascertain optical flow, necessitating the use of a supplementary constraint. For this, the smoothness constraint is often employed. This constraint is, however, general and does not express well a priori knowledge of a specific object. We therefore propose a method in which physical equations describing the object are used as supplementary constraints. In this way, more accurate flow estimation can be achieved. The physical model employed is a combination of the continuity equation and Navier-Stokes' equations. After describing how we integrate these equations into fluid flow estimation, we demonstrate the effectiveness of the proposed method by presenting experimental results of its application to simulated and real Karman flows. (C) 2002 Pattern Recognition Society. Published by Elsevier Science Ltd. All rights reserved.