A novel method for adaptive terrain rendering using memory-efficient tessellation codes for virtual globes

Abstract

Virtual globes render large-scale real-world earth’s surface in real-time for geographic information visualization. Modern GPUs have introduced hardware tessellation features for high-performance rendering. Patch-based terrain rendering using hardware tessellation introduces cracks and swimming artifacts during navigation. Hardware tessellation-based terrain rendering methods limit the tessellation factors to power-of-two and render patches with discrete-level-of-detail (DLOD). We present a novel method for adaptive terrain rendering using tessellation codes. In the proposed method, we present novel memory-efficient bit-field-based tessellation codes to encode patch Level-of-detail (LOD). The encoding scheme enabled patch rendering with Continuous-level-of-detail (CLOD) on tessellation hardware. The view-dependent simplification algorithm simplifies and encodes the patch’s Level-of-detail (LOD) in tessellation codes. The proposed method, based on tessellation codes, tessellates patches as uniformly spaced vertices grid. A novel relocation algorithm, based on tessellation codes, relocates uniformly spaced vertices to construct an encoded simplified vertices grid to avoid boundary cracks and swimming artifacts. The tessellation codes consumed 28 kilobytes of memory for a patch size of 65. The proposed method rendered 2.5 million triangles with a frame rate of 570 frames-per second for an ultra-HD display. The proposed method achieved a high simplification and rendering rate with low computational and memory usage as compared to state-of-the-art methods.

Keywords

• Real-time graphics
• Terrain modelling
• Level of detail
• Hardware tessellation
• Virtual globes