Parallelized 3D optical flow method for fluid mechanics applications

Tamas Regert (1, 2) , Benoit Tremblais (1) , Laurent David (2)

1 SIC
2 LEA - Laboratoire d'Études Aérodynamiques

In this paper we address the problem of estimating the motion of fluids in 3D image sequences. We present a 3D extension of the second order div-curl regularization 2D optical flow equation first introduced by Corpetti et al.. In addition, we propose a multigrid and Message Passing Interface (MPI) parallelized implementation of the algorithm to handle the huge amount of data encountered in our applications. The performance of the resulting fluid flow estimator is demonstrated on a well representative motion synthetic pattern called Hill's spherical vortex. The suggested method should be a good alternative or complement to the well established techniques based on 3D cross correlation.

Mots clés: image processing; 3d optical flow; PDE; MPI; multigrid

Domaines: Traitement du signal et de l'image [eess.SP]; Mécanique des fluides [physics.class-ph]; Mécanique des fluides [physics.class-ph]; Traitement du signal et de l'image [eess.SP]

Tamas Regert, Benoit Tremblais, Laurent David. Parallelized 3D optical flow method for fluid mechanics applications. FIFTH INTERNATIONAL SYMPOSIUM ON 3D DATA PROCESSING, VISUALIZATION AND TRANSMISSION, May 2010, Paris, France. pp.20. ⟨hal-00485126⟩