Feature-Preserving Displacement Mapping With Graphics Processing Unit (GPU) Tessellation

Abstract

Displacement mapping reconstructs a high-frequency surface by adding geometric details encoded in the displacement map to the coarse base surface. In the context of hardware tessellation supported by GPUs, this paper aims at feature-preserving surface reconstruction, and proposes the generation of a displacement map that displaces more vertices towards the higher-frequency feature parts of the target mesh. In order to generate the feature-preserving displacement map, surface features of the target mesh are estimated, and then the target mesh is parametrized and sampled using the features. At run time, the base surface is semi-uniformly tessellated by hardware, and then the vertices of the tessellated mesh are displaced non-uniformly along the 3-D vectors stored in the displacement map. The experimental results show that the surfaces reconstructed by the proposed method are of a higher quality than those reconstructed by other methods.