Duality TV-L1 flow with fundamental matrix prior

Variational techniques yield the most accurate results for dense optical flow fields between two images. They have the nice property of inherent smoothness to cope with untextured image regions: the filling-in of such regions is driven by neighbouring pixels. Such filling-in is not always the best choice. If the scene is mostly stationary and the camera is moving, the direction of the optical flow vectors can be restricted using the fundamental matrix. In this paper we propose an exact solution of the variational optical flow, using the fundamental matrix geometry as an additional weak prior. Our novel approach currently performs best on the Middlebury flow evaluation which includes images from stationary and dynamic scenes.