Digital Representations of the Real World

17. Image- and Video-Based Rendering 265

Figure 17.9: View-dependent texture mapping: if an accurate 3D geometry

model of the scene is available, it can be projectively textured using only a

few captured input images. Typically, more than one input image is used for

projective texturing to cover all visible regions of the object. The selection

of input images and their respective blending weights are assigned based

on the position of the viewpoint.

depth cameras or 3D scanners may be used to obtain partial object geome-

try [Wood et al. 00] (Chapter 9). Finally, a watertight 3D model is obtained

from the acquired point cloud using Poisson surface reconstruction ...