Dense Stereo Matching Based On Propagation with Voronoi Diagram.
Li Tang, Hung-Tat Tsui, Chengke Wu · Indian Conference on Computer Vision, Graphics and Image Processing · 2002
A new dense matching algorithm is proposed in this paper. It is based on propagation from N seed points, which have been matched reliably by feature tracking. The whole image is first divided into N cells by the Voronoi diagram of the seed feature points. Then corresponding relations are propagated from the seed in each cell until all pixels within this cell are processed. Modified sum of squared differences (SSD) is employed as the cost function in the propagation of matching according to a statistic model of disparity distribution within the window. The size of the window is adaptive. It is made inversely proportional to the texture density inside the window to increase the reliability of matching. A significant merit of the algorithm is that it can be applied to a wide range of image pairs including those with large disparities, with or without rectification. The algorithm has been verified with real images and the results show that it is both robust and accurate.