We have then considered data extracted from several experiments on the Coriolis rotating platform [50]. A large rotating turntable (diameter: 13 meters) allows us to reproduce the oceanic or atmospheric flows. Depending on the experiments, either some colorant or particles are inserted in the water as the platform rotates, and among the various measurement devices, a camera takes pictures of the experiment [62].
Several test cases have been studied, corresponding to either small or large acquisition times between two consecutive images. In all these different situations, the global structure of the displacement field matches perfectly with the real displacement of the fluid. The multi-grid approach has been compared with the standard approach, in which the minimization is directly performed on the fine grid. Both the computation time and the quality of the results are degraded.
These results have been compared with those produced by the PIV (Particle Imaging Velocimetry) method. PIV is the reference method for the extraction of velocity fields in geophysics and fluid mechancics. The results are qualitatively equivalent, in the sense that the identified fields look alike. However, our algorithm represents two main improvements: the computation time, allowing us to extract velocity fields from several hundreds of images in a relatively short time; and the preciseness of the results, as we extract one velocity vector for each pixel of the image, while the PIV method usually gives only one vector every nearly
pixels. This allows us to track the evolution of very small structures.