Physically based fluid flow recovery from image sequences

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. Results are presented from application of the algorithm to transmittance imagery of fluid flows, where the fluids contained a contrast medium. In these experiments, the algorithm recovered accurate and precise estimates of the flow. The significance of this work is two fold. 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.