Over-Parameterized Optical Flow Using a Stereoscopic Constraint [chapter]

Guy Rosman, Shachar Shem-Tov, David Bitton, Tal Nir, Gilad Adiv, Ron Kimmel, Arie Feuer, Alfred M. Bruckstein
2012 Lecture Notes in Computer Science  
The success of variational methods for optical flow computation lies in their ability to regularize the problem at the pixel level and combine piecewise smoothness of the flow field with brightness constancy assumptions. However, the piecewise smoothness assumption is often motivated by heuristic or algorithmic considerations. Lately, new priors were proposed to exploit the structural properties of the flow. Yet, most of them still utilize a generic diffusion term for regularization. Of
more » ... ar interest is the problem of optical flow estimation in static scenes for dense shape reconstruction. In this case, we show that introducing a motion model into the optical flow allows to adjust the smoothness term to scene priors. The proposed method assumes that the visible surface can be approximated by a piecewise smooth manifold. Accordingly, the optical flow between two consecutive frames can be locally regarded as a homography consistent with the epipolar geometry and defined by three parameters at each pixel. The resulting regularization term measures the total variation of the model parameters. This new technique yields significant improvements over state of the art optical flow computation methods for static scenes.
doi:10.1007/978-3-642-24785-9_64 fatcat:tjlwes3hfngjrlk27bckaxonhe