ABSTRACT
Virtual globes visualize massive quantities of terrain and imagery. Imagine a single rectangular image that covers the entire world with a sufficient resolution such that each square meter is represented by a pixel. The circumference of Earth at the equator and around the poles is roughly 40 million meters, so such an image would contain almost a quadrillion (1 × 1015) pixels. If each pixel is a 24-bit color, it would require over 2 million gigabytes of storage-approximately 2 petabytes! Lossy compression reduces this substantially, but nowhere near enough to fit into local storage, never mind on main or GPU memory, on today’s or tomorrow’s computers. Consider that popular virtual globe applications offer imagery at resolution higher than one meter per pixel in some areas of the globe, and it quickly becomes obvious that such a na¨ıve approach is unworkable.
