Volumetric Scene Reconstruction For View Synthesis

Bruce Culbertson, Tom Malzbender, John Harvey*

Visual Computing Dept., Hewlett-Packard Laboratories, Palo Alto, CA
*Department of Computer Science, Massachusetts Institute of Technology

 

Abstract

Our goal is to render new views from a set of captured images. The algorithms that we use require camera pose, which we automatically estimate with the Chromaglyph method. Voxel-coloring [Seitz Dyer 1997] is an algorithm for constructing a model of a scene, which can then be rendered for new viewpoints. We have developed several enhancements to voxel-coloring, plus a new related geometrical algorithm called Blinking Cubes.

Note: This is a Web version of a poster presented at ibr98, the IEEE Workshop on Image-based Modeling and Rendering, Stanford, CA, March 23-25, 1998. Postscript for the actual poster is available by clicking top and bottom.

Created 3 April 1998. For information, contact Bruce Culbertson.