🤖 AI Summary
This study addresses the challenge of identifying narrow structures—such as "necks"—on surfaces by proposing a formal definition of bottlenecks grounded in the isoperimetric inequality. Specifically, a bottleneck is characterized as a closed curve that partitions the surface and minimizes length relative to the enclosed area. Building on this theoretical foundation, the authors develop an efficient and practical algorithm for bottleneck detection and cutting, integrating tools from computational geometry and topology. The approach not only provides a rigorous theoretical basis for surface segmentation but also demonstrates significant improvements in segmentation quality and robustness across a variety of real-world datasets. Additional results and visualizations are available at https://neckcut.space.
📝 Abstract
We study the problem of finding neck-like features on a surface. Applications for such cuts include robotics, mesh segmentation, and algorithmic applications. We provide a new definition for a surface bottleneck -- informally, it is the shortest cycle relative to the size of the areas it separates. Inspired by the isoperimetric inequality, we formally define such optimal cuts, study their properties, and present several algorithms inspired by these ideas that work surprisingly well in practice. For examples of our algorithms, see https://neckcut.space.