A Statistical Regularization Framework for Estimating Normal Displacements along Contours with Subpixel Accuracy

In this paper we address the problem of computing normal displacements along contours in image sequences. It is well known that the aperture problem restricts the estimation to the only component that is locally reachable: the perpendicular-to-the-edge one. We model the moving edges as spatio-temporal surface patches in the image sequence space (x, y, t). A statistical regularization scheme based on Markov Random Fields allows us to get a homogeneous and reliable normal motion field along contours. Subpixel accuracy comes straight-forwardly with this modeling, and is handled within the optimization stage itself, not as a post-processing step. An efficient convolution-like implementation of the computations involved in the relaxation scheme is described. Results are presented concerning synthetic experiments and real-world sequences.