Efficient Evaluation and Smoothing of Continuous Signed Distance Fields for Polygonal Meshes
Authors: Sanchez, M., Fryazinov, O. and Pasko, A.
Editors: Rushmeier, H. and Deussen, O.
Journal: Computer Graphics Forum: the international journal of the Eurographics Association
Abstract:Signed distance fields are commonly used for representing polygonal meshes in various applications. However, they introduce $C^1$-discontinuities across the field causing creases to appear when applying operations such as blending and metamorphosis. The focus of this work is to efficiently evaluate signed distance function and smooth out the distance field. We compare several approaches to evaluating the signed distance and choose the most efficient one. The details of a GPU implementation are also discussed and packet queries are introduced to support several common operations on field values. Finally, the field is smoothed out using a discrete filter over the continuous function. The resulting function provides a numerically $C^1$-continuous field which preserves the exact shape of the polygonal mesh. Thanks to its low complexity, the proposed filtering technique is fast compared to its main alternatives providing $C^1$-continuous distance field approximations. Several application examples are presented such as blending set operations, linear metamorphosis, space-time blending and microstructure generation for polygonal meshes.
Source: Manual