QuickCSG: Fast Arbitrary Boolean Combinations of N Solids

TL;DR

QuickCSG introduces a vertex-centric, KD-tree based algorithm for fast boolean operations on multiple polyhedra, achieving significant speedups over existing CPU and GPU methods while maintaining output quality.

Contribution

The paper presents a novel vertex-centric approach combined with spatial decomposition and parallelization, enabling efficient boolean operations on multiple solids without intermediate trees.

Findings

Achieves 10-100x speedups over state-of-the-art CPU algorithms.

Outperforms GPU implementations with approximate discretizations.

Effective for large CSG problems in applications like 3D printing and collision detection.

Abstract

QuickCSG computes the result for general N-polyhedron boolean expressions without an intermediate tree of solids. We propose a vertex-centric view of the problem, which simplifies the identification of final geometric contributions, and facilitates its spatial decomposition. The problem is then cast in a single KD-tree exploration, geared toward the result by early pruning of any region of space not contributing to the final surface. We assume strong regularity properties on the input meshes and that they are in general position. This simplifying assumption, in combination with our vertex-centric approach, improves the speed of the approach. Complemented with a task-stealing parallelization, the algorithm achieves breakthrough performance, one to two orders of magnitude speedups with respect to state-of-the-art CPU algorithms, on boolean operations over two to dozens of polyhedra. The algorithm also outperforms GPU implementations with approximate discretizations, while producing an output without redundant facets. Despite the restrictive assumptions on the input, we show the usefulness of QuickCSG for applications with large CSG problems and strong temporal constraints, e.g. modeling for 3D printers, reconstruction from visual hulls and collision detection.