Photorealistic terrain visualization results from combining two data sets. The first contains information about terrain color (texture), usually from a vertical view angle, such as an aerial or satellite image. The second data set contains information about terrain topography, in the form of elevation samples. This data set is also known as a digital terrain model, or DTM. We can reconstruct the 3D terrain from the DTM using various methods. The conventional approach triangulates the terrain into a continuous surface consisting of relatively large planar facets. An alternative approach uses a regular array of atomic values called voxels to represent the terrain. After terrain surface reconstruction, we render the oblique perspective terrain image by a process called phototexturing-the mapping of the corresponding texture onto this surface. Before doing this, we must register the texture and DTM so that the overlay is accurate. This corrects geometric distortions in the data sets, which originate in measurement device or sensor inaccuracies. We obtain the final image by projecting the colored surface onto a viewing plane, incorporating hidden surface elimination. © 1994 IEEE.
