SMP stereo matching based on shift-invariant NSCT

One of the central problems in stereo matching is the selection of optimal window sizes for comparing image regions. This paper presents a fast and reliable multi-scale stereo matching algorithm based on the Nonsubsampled Contourelet Transform (NSCT), which is a fully shift-invariant, multi-scale, and multi-direction expansion that has better directional frequency localization. This method consists of multiple passes which compute stereo matches with a coarse-to-fine and sparse-to-dense paradigm. Then, we using a fast local algorithm, referred to as single matching phase (SMP), which reduce computation and is suitable to real time applications. The proposed approach is also compared with bidirectional matching (BM), since the latter is the basic method for detecting unreliable matches in most area-based stereo algorithms. Experimental results with real images are presented.