Geospecific rendering of terrain requires information about both the geometry and the photometry of the scene. Raw information about the geometric shape of the terrain itself is most often available as a Digital Elevation Model (DEM), in which elevation values are represented in a rectangular grid. The highest resolution widely available elevation data for the continental U.S. are United States Geological Survey (USGS) 7.5-Minute DEMs [22]. Elevation values with a nominal precision of 1m are provided at 30m intervals (post spacing) on the ground. The 7.5-Minute DEMs are created by optically scanning contour maps and then fitting an approximation surface. They are subject to a number of systematic distortions that, depending on the technology used when a particular DEM was produced, can result in the actual resolvability of ground features being far worse than the 30m post spacing might suggest.
The most direct way to render geospecific photometry is to start with an image of the area to be rendered. Perspective effects make it difficult to register conventional aerial imagery with elevation data. As a result, an orthorectification process is often performed, in which the perspective image is warped to remove the effects of lens projection, camera orientation, and terrain. The result is an image that is effectively a scaled orthographic projection. The USGS provides 1m resolution panchromatic (greyscale) Digital Orthoimagery (DOQ) for much of the continental U.S. [21]. No comparable source for color orthoimagery exists. Aerial survey companies can produce such imagery on a custom basis, but the cost is significant. Satellite images are often used to render terrain. While not true color imagery as that term is commonly used, multi-spectral satellite data can be converted to an RGB format that closely approximates perceptual color. In addition, much work has gone into the classification of multi-spectral satellite data to determine properties such as vegetation cover. Unfortunately, the resolution of available multi-spectral satellite data is at best on the order of 20m on the ground [1,18].
If actual imagery of the terrain being rendered is to be used, the available choices are usually limited to false-colored multi-spectral satellite data of limited resolution or USGS high-resolution panchromatic orthoimagery. Since visual realism in terrain rendering depends in part on high resolution texturing, it is important to explore whether or not panchromatic orthoimagery can be effectively utilized. Can realistic color be generated? Can we get enough information about ground cover to add detail not resolved in the imagery? Can we remove shadows and shading effects so as to simulate views at times other than when the original imagery was acquired? Can we simulate seasonal effects not present in the original imagery?