Depth Discontinuities by Pixel-to-Pixel Stereo

  • Authors:
  • Stan Birchfield;Carlo Tomasi

  • Affiliations:
  • Department of Computer Science, Stanford University, Stanford, CA 94305. birchfield@cs.stanford.edu;Department of Computer Science, Stanford University, Stanford, CA 94305. tomasi@cs.stanford.edu

  • Venue:
  • International Journal of Computer Vision
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

An algorithm to detect depth discontinuities from a stereo pair ofimages is presented. The algorithm matches individual pixels incorresponding scanline pairs, while allowing occluded pixels toremain unmatched, then propagates the information betweenscanlines by means of a fast postprocessor. The algorithm handleslarge untextured regions, uses a measure of pixel dissimilaritythat is insensitive to image sampling, and prunes bad search nodesto increase the speed of dynamic programming. The computation isrelatively fast, taking about 600 nanoseconds per pixel perdisparity on a personal computer. Approximate disparity maps andprecise depth discontinuities (along both horizontal and verticalboundaries) are shown for several stereo image pairs containingtextured, untextured, fronto-parallel, and slanted objects inindoor and outdoor scenes.