Variational optic flow computation in real-time

Abstract

Variational methods for optic flow computation have the reputation of producing good results at the expense of being too slow for realtime applications. We show that real-time variational computation of optic flow fields is possible when appropriate methods are combined with modern numerical techniques. We consider the CLG method, a recent variational technique that combines the quality of the dense flow fields of the Horn and Schunck approach with the noise robustness of the Lucas-Kanade method. For the linear system of equations resulting from the discretised Euler-Lagrange equations, we present different multigrid schemes in detail. We show that under realistic accuracy requirements they are up to 247 times more efficient than the widely used Gauß-Seidel algorithm. On a 3.06 GHz PC, we have computed 40 dense flow fields of size 200 x 200 pixels within a single second.