Recovering estimates of fluid flow from image sequence data

This paper presents an approach to measuring fluid flow from image sequences. The approach centers around a motion-recovery algorithm that is based on principles from fluid mechanics: The algorithm is constrained so that recovered flows observe conservation of mass as well as physically motivated boundary conditions. Empirical results from application of the algorithm to transmittance imagery of fluid flows, where the fluids contained a contrast medium, are presented. In these experiments, the algorithm recovered accurate and precise estimates of the flow. The significance of this work is twofold. First, from a theoretical point of view it is shown how information derived from the physical behavior of fluids can be used to motivate a flow-recovery algorithm. Second, from an applications point of view the developed algorithm can be used to augment the tools that are available for the measurement of fluid dynamics; other imaged flows that observe compatible constraints might benefit in a similar fashion. (C) 2000 Academic Press.