Voting for Distortion Points in Geometric Processing

Shuangming Chai, Xiao-Ming Fu, Ligang Liu

Abstract

Low isometric distortion is often required for mesh parameterizations. A configuration of some vertices, where the distortion is concentrated, provides a way to mitigate isometric distortion, but determining the number and placement of these vertices is non-trivial. We call these vertices distortion points. We present a novel and automatic method to detect distortion points using a voting strategy. Our method integrates two components: candidate generation and candidate voting. Given a closed triangular mesh, we generate candidate distortion points by executing a three-step procedure repeatedly: (1) randomly cut an input to a disk topology; (2) compute a low conformal distortion parameterization; and (3) detect the distortion points. Finally, we count the candidate points and generate the final distortion points by voting. We demonstrate that our algorithm succeeds when employed on various closed meshes with a genus of zero or higher. The distortion points generated by our method are utilized in three applications, including planar parameterization, semi-automatic landmark correspondence, and isotropic remeshing. Compared to other state-of-the-art methods, our method demonstrates stronger practical robustness in distortion point detection.

Introduction

In geometric processing, parameterization is the process of flattening a 3D surface onto a 2D plane—a task analogous to peeling an orange. Because most surfaces are not "developable" (like a cylinder or cone), flattening them inevitably introduces distortion in the form of stretching or squashing. To minimize this isometric distortion, researchers often introduce cuts into the mesh. The quality of the final flattening depends heavily on whether these cuts pass through "distortion points"—vertices where the surface's curvature or topology makes flattening particularly difficult.

While humans can often identify these points intuitively, automatic detection is challenging. Existing methods frequently struggle with complex, thin geometries or require computationally expensive spherical mappings. This paper introduces a robust, automatic voting-based strategy to identify these critical vertices.

Methodology

The proposed framework operates in two primary stages: Candidate Generation and Candidate Voting.

  1. Candidate Generation: The algorithm repeatedly performs a three-step loop (typically 10 iterations). First, the closed mesh is randomly cut to create a disk topology. Second, the mesh is flattened using a "flip-free" conformal mapping to minimize angle distortion. Finally, areas of high isometric distortion are clustered to identify candidate vertices.
  2. Voting Strategy: A single random cut might create "fake" distortion artifacts near the cut line. To filter this noise, the algorithm uses a voting strategy. Vertices that consistently appear as candidates across multiple different random cuts are promoted to the final set of distortion points. This leverages the instability of individual cuts to find the underlying geometric features that remain consistent.
  3. Refinements: To ensure efficiency and precision, the authors utilize Mesh Simplification (QEM) to speed up processing on high-resolution models and apply post-filtering to group nearby candidate points into a single representative vertex.

Key Results and Performance

The voting-based method consistently produces lower average ($\delta_{avg}$) and maximum ($\delta_{max}$) isometric distortion compared to state-of-the-art methods like Seamster or Geometry Images. For instance, on a high-resolution armchair model, the curvature-based Seamster method finds redundant points and misses important features, leading to a $\delta_{max}$ of 12.6. In contrast, the proposed method correctly identifies features like the backrest, achieving a significantly lower $\delta_{max}$ of 3.75.

Comparison of distortion point detection and resulting isometric distortion between state-of-the-art methods and the proposed voting strategy.

Efficiency is another key advantage. By utilizing mesh simplification, the algorithm can process complex models, such as an 83k-vertex Armadillo, in approximately 20 seconds—more than twice as fast as processing the full-resolution mesh directly.

Practical Applications

The distortion points generated by this method serve as a robust foundation for several downstream geometric tasks:

  • Planar Parameterization: By cutting through the detected points, the algorithm achieves high-quality UV mappings with minimal stretching.
  • Surface Correspondence: The detected points serve as excellent anchors for mapping one 3D shape onto another, facilitating semi-automatic landmark correspondence between different models (e.g., two different bunny or fish meshes).
  • Isotropic Remeshing: The method successfully preserves thin features, such as a dog's tail or a fish's fins, which often collapse during standard remeshing processes.
Applications of the detected distortion points in planar parameterization, surface correspondence, and isotropic remeshing.

Conclusion and Limitations

While the voting strategy demonstrates strong practical robustness, it remains an empirical method without a formal theoretical proof linking the voting threshold to guaranteed distortion bounds. Additionally, the method may struggle with perfectly symmetrical objects like spheres, where every vertex is an equally likely candidate, and requires pre-processing (hole filling) for high-genus meshes. Despite these limitations, the approach offers a significant advancement in automatic mesh processing for diverse and complex geometries.

Summary updated

Abstract

Low isometric distortion is often required for mesh parameterizations. A configuration of some vertices, where the distortion is concentrated, provides a way to mitigate isometric distortion, but determining the number and placement of these vertices is non-trivial. We call these vertices distortion points. We present a novel and automatic method to detect distortion points using a voting strategy. Our method integrates two components: candidate generation and candidate voting. Given a closed triangular mesh, we generate candidate distortion points by executing a three-step procedure repeatedly: (1) randomly cut an input to a disk topology; (2) compute a low conformal distortion parameterization; and (3) detect the distortion points. Finally, we count the candidate points and generate the final distortion points by voting. We demonstrate that our algorithm succeeds when employed on various closed meshes with a genus of zero or higher. The distortion points generated by our method are utilized in three applications, including planar parameterization, semi-automatic landmark correspondence, and isotropic remeshing. Compared to other state-of-the-art methods, our method demonstrates stronger practical robustness in distortion point detection.
BibTeX
@article{chai2019voting,
  author = {Shuangming Chai and Xiao-Ming Fu and Ligang Liu},
  title = {{Voting for Distortion Points in Geometric Processing}},
  journal = {IEEE Transactions on Visualization and Computer Graphics},
  year = {2019},
  doi = {10.1109/tvcg.2019.2947420},
}
Figure