Why Wavelets for Terrain Storage?
Virtually all authored terrain datasets originate as regular-gridded raster arrays of elevation heights (e.g., DTED®s, DEMs, DTMs). Wavelets are extremely well suited for compression of raster arrays and inherently provide powers-of-2 levels-of-detail (LOD) reconstruction capabilities. Additionally, arbitrary blocks of terrain data at any arbitrary LOD can be efficiently extracted in real time from a wavelet-compressed terrain dataset.

TINing Versus Wavelets.
TINing approaches have been used as a method of vertex thinning—and correspondingly, polygon count reduction—to achieve increased terrain-rendering display rates. The problem with TINed terrain models is that errors are created in the models at author time, and vertices within the models must use world-space floating point coordinates because of their spatially irregular nature. The typical result is a single-LOD terrain model with built-in errors and absolute coordinates for every vertex. As terrain roughness increases or model error requirements decrease, TIN model sizes increase rapidly. TerraBlocks™ terrain data wavelet-compression maintains controlled data size regardless of terrain roughness, error requirements or the number of available LODs. All vertex coordinates are stored as small local-offset data types.

TerraBlocks™ Terrain-Block Rendering.
Individual TerraBlocks™ are extracted at render time and processed as packets. Blocks are accessed and processed at viewpoint-appropriate LODs. Block seaming and parent-child interpolation provide smooth, continuous LOD vertex transitioning, which eliminates vertex popping. In the final render cycle, TerraBlocks™ are placed “on the sphere” relative to the viewpoint. This makes efficient use of the render pipeline, provides on-the-sphere rendering and produces zero visual rendering artifacts. TerraBlocks™ makes applications such as synthetic vision for approach and landing, and low-level, zero-visibility flight a reality for commercial transport and military applications.