Fast and Accurate Large-Scale Stereo Reconstruction Using Variational Methods

This paper presents a fast algorithm for high-accuracy large-scale outdoor dense stereo reconstruction of man-made environments. To this end, we propose a structure-adaptive second-order Total Generalized Variation (TGV) regularization which facilitates the emergence of planar structures by enhancing the discontinuities along building facades. As data term we use cost functions which are robust to illumination changes arising in real world scenarios. Instead of solving the arising optimization problem by a coarse-to-fine approach, we propose a quadratic relaxation approach which is solved by an augmented Lagrangian method. This technique allows for capturing large displacements and fine structures simultaneously. Experiments show that the proposed augmented Lagrangian formulation leads to a speedup by about a factor of 2. The brightness-adaptive second-order regularization produces sub-disparity accurate and piecewise planar solutions, favoring not only fronto-parallel, but also slanted planes aligned with brightness edges in the resulting disparity maps. The algorithm is evaluated and shown to produce consistently good results for various data sets (close range indoor, ground based outdoor, aerial imagery).