Traditionally, the problem of stereo matching has been addressed either by a local window-based approach or a dense pixel-based approach using global optimization. In this paper, we present an algorithm which combines window-based local matching into a global optimization framework. Our local matching algorithm assumes that local windows can have at most two disparities. Under this assumption, the local matching can be performed very efficiently using graph cuts. The global matching is formulated as minimization of an energy term that takes into account the matching constraints induced by the local stereo algorithm. Fast, approximate minimization of this energy is achieved through graph cuts. The key feature of our algorithm is that it preserves discontinuities both during the local as well as global matching phase.
Motilal Agrawal, Larry S. Davis