Stereo Matching with Transparency and Matting

  • Authors:
  • Richard Szeliski;Polina Golland

  • Affiliations:
  • Microsoft Reasearch, One Microsoft Way, Redmond, WA 98052-6399. szeliski@microsoft.com;Artificial Intelligence Laboratory, 545 Technology Square \'#810, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139. polina@ai.mit.edu

  • Venue:
  • International Journal of Computer Vision - 1998 Marr Prize
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

This paper formulates and solves a new variant of the stereocorrespondence problem: simultaneously recovering the disparities,true colors, and opacities of visible surface elements. This problemarises in newer applications of stereo reconstruction, such as viewinterpolation and the layering of real imagery with synthetic graphicsfor special effects and virtual studio applications. While this problemis intrinsically more difficult than traditional stereo correspondence,where only the disparities are being recovered, it provides a principledway of dealing with commonly occurring problems such as occlusions andthe handling of mixed (foreground/background) pixels near depthdiscontinuities. It also provides a novel means for separatingforeground and background objects (matting), without the use of a specialblue screen. We formulate the problem as the recovery of colors andopacities in a generalized 3D (x, y, d) disparity space, and solve theproblem using a combination of initial evidence aggregation followed byiterative energy minimization.