Coarse-to-Fine Stereo with Accurate 3D Boundaries

Contributors

• Mikhail Sizintsev
• Richard P. Wildes

Abstract:

This research develops methods for efficient recovery of accurate binocular disparity estimates in the vicinity of 3D surface discontinuities. Of particular concern are methods that impact coarse-to-fine, local block-based matching, as it forms the basis of the fastest and most resource efficient stereo computation procedures. A novel coarse-to-fine refinement procedure that adapts match window support across scale to amelieorate corrupt disparity estimation near boundaries is presented. Extensions are included to account for half-occlusions and colour uniformity. Empirical results show that incorporation of these advances in the standard coarse-to-fine, block matching framework reduces disparity errors by more than a factor of two, while performing little extra computation, preserving low computational complexity and the parallel/pipeline nature of the paradigm. Moreover, the proposed advances are shown to be beneficial to coarse-to-fine global matchers as well.

Related Publications:

• M. Sizintsev and R. Wildes, Coarse-to-fine stereo vision with accurate 3D boundaries, Image and Vision Computing (IVC), 28, pp. 352-366, 2010.
• M. Sizintsev and R. P. Wildes, Computational analysis of binocular half-occlusions. In L. Harris and M. Jenkin (Eds.) Computational Vision in Neural and Machine Systems, Cambrdige University Press: Cambridge, UK, 2007. (See earlier.)
• M. Sizintsev and R.P. Wildes, Coarse-to-Fine Stereo Vision with Accurate 3-D Boundaries York University Technical Report CS-2006-07, June 28, 2006
• M. Sizintsev and R.P. Wildes, Efficient stereo with accurate 3-D boundaries. In Proceedings of the British Machine Vision Conference (BMVC), pp. 237-246, 2006.