Feature-Preserving Cage Construction based on Signed Distance Field.

Area of Science:

• Computer Graphics
• Computational Geometry
• Geometric Modeling

Background:

• Generating simplified geometric representations is crucial for various applications.
• Existing methods often struggle with preserving intricate features and enforcing symmetry simultaneously.
• Automatic construction of high-quality, low-complexity cages remains a significant challenge.

Purpose of the Study:

• To develop a novel automatic approach for constructing low-complexity, manifold, and non-self-intersecting cages.
• To preserve geometric features, maintain tight enclosure, enforce symmetry, and produce high-quality meshes.
• To outperform existing techniques in handling complex shapes and generating compact, coarse cages.

Main Methods:

• A three-stage pipeline involving isosurface extraction from Signed Distance Fields (SDF), constrained Quadric Error Metric (QEM) simplification using a push-pull strategy, and a symmetry-aware refinement stage.
• Leveraging sharp features for shape-preserving initial cage construction and density field modification for non-self-intersection properties.
• Employing a symmetry deviation metric (SDE) for symmetry plane extraction and integrated optimization for mesh quality enhancement and collision-free enforcement.

Main Results:

• The proposed method successfully constructs low-complexity, manifold, and non-self-intersecting cages.
• Geometric features and symmetry are automatically preserved, with tight enclosure and high mesh quality.
• The method demonstrates superior performance compared to existing techniques on diverse models and public benchmarks.

Conclusions:

• The novel automatic approach effectively generates compact, coarse cages while preserving essential geometric properties and symmetry.
• This method offers a robust solution for complex shape simplification and feature-aware cage construction.
• The pipeline provides a significant advancement in automated geometric modeling and mesh generation.