Continuous Stereo Matching using Local Expansion Moves

We present an accurate and efficient stereo matching method using local expansion moves, a new move making scheme using graph cuts. The local expansion moves are presented as many alpha-expansions defined for small grid regions. The local expansion moves extend the traditional expansion moves by two ways: localization and spatial propagation. By localization, we use different candidate alpha-labels according to the locations of local alpha-expansions. By spatial propagation, we design our local alpha-expansions to propagate currently assigned labels for nearby regions. With this localization and spatial propagation, our method can efficiently infer Markov random field models with a huge or continuous label space using a randomized search scheme. Our local expansion move method has several advantages over previous approaches that are based on fusion moves or belief propagation; it produces submodular moves deriving a subproblem optimality; it helps find good, smooth, piecewise linear disparity maps; it is suitable for parallelization; it can use cost-volume filtering techniques for accelerating the matching cost computations. Our method is evaluated using the Middlebury stereo benchmark and shown to have the best performance in sub-pixel accuracy.