Motion Estimation in Computer Vision: Optimization on Stiefel Manifolds
نویسندگان
چکیده
Motion recovery from image correspondences is typically a problem of optimizing an objective function associated with the epipolar (or LonguetHiggins) constraint. This objective function is defined on the so called essential manifold. In this paper, the intrinsic Riemannian structure of the essential manifold is thoroughly studied. Based on existing optimization techniques on Riemannian manifolds, in particular on Stiefel manifolds, we propose a Riemannian Newton algorithm to solve the motion recovery problem, making use of the natural geometric structure of the essential nianifold. Although only the Newton algorithm is studied in detail, the same ideas also apply to other typical conjugate gradient algorithms. It is shown that the proposed nonlinear algorithms converge very rapidly (with quadratic rate of convergence) as long as the conventional SVD based eight-point linear algoritlini has a unique solution. Such Riemannian algorithms have also been applied to the differential (or continuous) case where the velocities are recovered from optical flows.
منابع مشابه
Advances in matrix manifolds for computer vision
This paper presents an overview of various matrix manifolds that are commonly used in computer vision applications. It covers the following manifoldsLie groups, Stiefel manifolds, Grassmann manifolds and Riemannian manifolds. A manifold of dimension n is a topological space that near each point resembles n-dimensional Euclidean space. More precisely, each point of an n-dimensional manifold has ...
متن کاملOptimal Motion Estimation from Multiple Images by Normalized Epipolar Constraint
In this paper, we study the structure from motion problem as a constrained nonlinear least squares problem which minimizes the so called reprojection error subject to all constraints among multiple images. By converting this constrained optimization problem to an unconstrained one, we contend that multilinear constraints, when used for motion and structure estimation, need to be properly normal...
متن کاملGeneralized BackPropagation, Étude De Cas: Orthogonality
This paper introduces an extension of the backpropagation algorithm that enables us to have layers with constrained weights in a deep network. In particular, we make use of the Riemannian geometry and optimization techniques on matrix manifolds to step outside of normal practice in training deep networks, equipping the network with structures such as orthogonality or positive definiteness. Base...
متن کاملGlobal optimization on Stiefel manifolds — some particular problem instances ∗
Optimization on Stiefel manifolds was discussed by Rapcsák in earlier papers. There, some numerical methods of global optimization are dealt with and tested on Stiefel manifolds. In the paper the structure of the optimizer points is given in some particular problem instances and for a special form of a quadratic problem defined on a Stiefel manifold. Some reduction tricks and results are obtain...
متن کاملOptimal Motion Estimation from Multiview Normalized Epipolar Constraint
In this paper, we study the structure from motion problem as a constrained nonlinear least squares problem which minimizes the so called reprojection error subject to all constraints among multiple images. By converting this constrained optimization problem to an unconstrained one, we obtain a multiview version of the normalized epipolar constraint of two views. Such a multiview normalized epip...
متن کامل