This clearly written and well-organized paper examines the utility of Delaunay triangulation for an object-centered representation of three-dimensional objects for which multiple views are available. This approach marks empty tetrahedra to deal with visibility and yields both surface representations of objects and volumetric representations of free space. The authors demonstrate that a constrained Delaunay triangulation is easy to update, able to accommodate both sparse and dense data, and computationally efficient for the interpolation of stereo data. Computational efficiency depends, of course, on the representativeness of skeletal approximations of the object’s interior and exterior surfaces. An analysis of complexity for average-case and worst-case situations accompanies a lucid, well-illustrated description of several relevant image-processing algorithms. The authors provide a critical appraisal of the utility of Delaunay triangulation for stereo interpolation and relate their work to previous attempts to deal with the problem. Although the paper is illustrated with numerous examples of room-scale, camera-based line-geometry stereo data, the Delaunay-triangulation approach preserves depth discontinuities and thus might be useful for data captured by sonar and other active remote sensing systems, including airborne terrain sensing systems.