Optical flow estimation based on the extraction of motion patterns

In this paper, a new methodology for optical flow estimation that is able to represent multiple motions is presented. To separate motions at the same location, a new frequency-domain approach is used. This model, based on a band-pass filtering with a set of log-Gabor spatio-temporal filters, groups together filter responses with continuity in its motion (each group will define a motion pattern). Given a motion pattern, the gradient constraints is applied to the output of each filter in order to obtain multiple estimates of the velocity at the same location. Then, the velocities at each point of the motion pattern are combined using probabilistic rules. The use of "motion patterns" allows to represent multiple motions, while the combination of estimates from different filters helps to reduce the initial aperture problem. This technique is illustrated on real and simulated data sets, including sequences with occlusion and transparencies.