Robust generation of signed distance fields from triangle meshes

Abstract

A new method for robust generation of distance fields from triangle meshes is presented. Graphics hardware is used to accelerate a technique for generating layered depth images. From multiple layered depth images, a binary volume and a point representation are extracted. The point information is then used to convert the binary volume into a distance field. The method is robust and handles holes, spurious triangles and ambiguities. Moreover, the method lends itself to Boolean operations between solids. Since a point cloud as well as a signed distance is generated, it is possible to extract an iso-surface of the distance field and fit it to the point set. Using this method, one may recover sharp edge information. Examples are given where the method for generating distance fields coupled with mesh fitting is used to perform Boolean and morphological operations on triangle meshes.