Decoding of Large Terrains Using a Hardware Rendering Pipeline †

In this paper, we present a simple approach to the quantization of vertex-coordinate values of a very large terrain dataset, allowing both efficient representation and timely rendering of the data. We focus on uniform scalar quantization, which, although theoretically suboptimal in general, is shown to produce the same level of accuracy as a Lloyd-Max optimal quantizer for the terrain data under consideration. We find that a rate of quantization on the order of 6 bits per vertex results in visually lossless rendering of a very large terrain, allowing fly-throughs of high quality at interactive frame rates. Our novel approach to uniform scalar quantizer represents the inverse-quantization operation as a homogeneous matrix, allowing the rendering of the terrain data directly from the encoded representation with inverse quantization taking place automatically in the hardware rendering pipeline without the assumption of any special decompression engine. Additionally, our formulation is general and therefore amenable to incorporation in more sophisticated geometric-compression systems.