Incorporating discontinuity preserving approach for optical flow estimation

摘要

In this paper, we consider a modified variational energy functional for estimating optical flow from a sequence of images. The given functional incorporates the global variational model of Horn and Schunck(1981) and the classical model of Nagel and Enkelmann(1986) as a new regularisation functional, which efficiently preserve discontinuity in optical flow. In particular, the other significance of this work is in two aspects: (i) It uses a new diffusion model to estimate average velocities, which reflects the effect of the intensity difference. (ii) It uses a more efficient compensating iterative numerical scheme to achieve higher computational efficiency in order to solve the related Euler-Lagrange equations. Experimental evaluation on a given data sets (synthetic and real) present the robustness and accuracy of method and also validate the effect on discontinuity preserving.