A nonlinear fractional order variational model for the robust estimation of optical flow

This paper introduces a nonlinear fractional order variational model for the robust estimation of optical flow. In particular, it can be used to generalize the existing models from integer to fractional order. The proposed model combines a nonlinear Geman-McClure norm based data term with a Marchaud fractional derivative based regularization term. It significantly offers robustness against outliers, efficiently handle discontinuous texture and edge information and provides a dense flow field. The numerical discretization of the variational functional is facilitated using the Gru' nwald-Letnikov fractional derivative. Through an outer fixed point iteration scheme, the discretized nonlinear system is transformed into a linear system of equations, which is further solved using iterative techniques. Experimental evaluations across diverse datasets assess the model performance using various error measures (AAE, AEE, AENG). The results are also compared with the state-of-the-art-techniques.