Singular Arrange and Traverse Algorithm for Computing Reeb Spaces of Bivariate PL Maps

TL;DR

This paper introduces a new exact and efficient algorithm for computing the Reeb space of bivariate PL maps, significantly improving performance over previous methods and enabling practical applications in visualization and shape analysis.

Contribution

The paper presents the singular arrange and traverse algorithm, exploiting the fact that only singular edges affect the Reeb space, leading to substantial efficiency improvements.

Findings

Performance gains of up to four orders of magnitude on real datasets

Efficient computation of Reeb spaces for practical applications

Implementation demonstrates practical feasibility and advantages

Abstract

We present an exact and efficient algorithm for computing the Reeb space of a bivariate PL map. The Reeb space is a topological structure that generalizes the Reeb graph to the setting of multiple scalar-valued functions defined over a shared domain, a situation that frequently arises in practical applications. While the Reeb graph has become a standard tool in computer graphics, shape analysis, and scientific visualization, the Reeb space is still in the early stages of adoption. Although several algorithms for computing the Reeb space have been proposed, none offer an implementation that is both exact and efficient, which has substantially limited its practical use. To address this gap, we introduce singular arrange and traverse, a new algorithm built upon the arrange and traverse framework. Our method exploits the fact that, in the bivariate case, only singular edges contribute to the structure of Reeb space, allowing us to ignore many regular edges. This observation results in substantial efficiency gains on datasets where most edges are regular, which is common in many numerical simulations of physical systems. We provide an implementation of our method and benchmark it against the original arrange and traverse algorithm, showing performance gains of up to four orders of magnitude on real-world datasets.