Efficient terrain data coding for rendering on mobile devices

Abstract

Efficient terrain coding is necessary for large terrain real-time rendering on mobile devices. This paper is devoted to developing terrain compression algorithms so that large terrain can be rendered in real-time on mobile devices based on in-core memory. It presents lightweight compression techniques, namely binLBT+SPIHT and SPITH for efficient terrain digital elevation model (DEM) coding and shows the applications of ETC2 method for terrain texture coding for rendering on mobile devices. For DEM, terrain residual data is obtained and compressed in an integer lapped biorthogonal transform (binLBT) with lifting scheme followed by SPIHT (set partitioning in hierarchical trees) coder, or the terrain residual data is coded by SPIHT directly. For terrain texture, texture blocks are obtained by dividing the original large texture image, and each texture block is compressed by ETC2. Compression performances and reconstruction errors are compared in different parameters for binLBT+SPIHT, SPIHT and more wavelet coders. Based on these algorithms, we developed an interactive browse system of terrain data on mobile devices, using OpenGL ES 3.0, C++, NDK and so on. The tests on Puget Sound Terrain data with size 16384×16384 show that rendering achieves about 50 frames/sec.