Joint Estimation-Segmentation of Optic Flow

Abstract

In this paper we address the intricate issue of jointly recovering the apparent velocity field between two consecutive frames and its underlying partition. We design a global cost functional including robust estimators. These estimators enable to deal with the large deviations occurring in the different energy terms and offer the possibility to introduce a simple coupling between a dense optical flow field and a segmentation. This coupling is also reinforced by a parametric likeness term. The resulting estimation-segmentation model thus involves a tight cooperation between a local estimation process and a global modelization. The minimization of the final cost function is conducted efficiently by a multigrid optimization algorithm.