arXiv Science⌕ Search

arXiv · 2611.07328

DC-SQEM: Medial Mesh Extraction by Dual Contouring with Spherical Quadric Error Metric

Abstract

The medial axis, as a compact representation, describes not only the boundary but also the interior of the shape, and has been widely used in analyzing and simplifying shapes. However, approximating feature-preserving medial axis from unstructured data, such as oriented point clouds, remains challenging due to the lack of explicit connectivity and the presence of noise in the source. We propose DC-SQEM, a dual-contouring-based framework that computes medial meshes (an explicit representation of the medial axis) directly from oriented point clouds without relying on watertight manifold surfaces or volumetric structures. With the help of an adaptive octree, our method fits local medial spheres for each cell based on the Spherical Quadric Error Metric (SQEM). We address several long-standing robustness issues in dual-contouring and mesh extraction, including out-of-cell solutions, unstable feature classification, and redundant topological connections. To this end, we introduce virtual nodes, a feature-downgrading strategy, and a geometry-driven topology-aware cleanup procedure, enabling robust SQEM-based medial dual contouring with clean mesh outputs. Unlike prior point-cloud-based methods that produce either only 1D skeletal structures or incomplete medial features, our method preserves sharp geometric features from the input and reconstructs a complete medial structure, including sheets, curves, seams, and junctions. Compared to restricted-power-diagram-based approaches, our algorithm can handle point clouds as input, without having a strong requirement on water-tightness or manifoldness.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Bingchuan Li, Ruilong Liu, Tianhao Hu, Congyi Zhang, Ningna Wang, Junfeng Yao, Xiaohu Guo. 2026-11-06. DC-SQEM: Medial Mesh Extraction by Dual Contouring with Spherical Quadric Error Metric. https://arxiv.org/abs/2611.07328

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

Refining and Repairing Signed Distance Data With Global Consistency

We propose a method to interpolate exact Signed Distance Function (SDF) data from a discrete set of samples. Unlike prior work, our approach ensures that the new SDF data values are fully consistent with the input and each other, such that the augmented data still corresponds to a geometrically realizable surface. We express the theoretical properties of exact SDFs as hard geometric constraints, and construct an efficient greedy algorithm for consistent SDF interpolation that is made even faster with powerful parallelized GPU preprocessing. We exemplify the usefulness of our method by evaluating it on three practical applications: global SDF refinement, in which the SDF data is upsampled without knowledge of the ground truth; mesh reconstruction, where our method can reconstruct highly detailed surfaces using global information from coarse input SDFs; and repair of pseudo-SDFs, which result from many pipelines such as CSG Boolean operations and must be turned into valid SDFs for downstream processing tasks. Our refined SDFs are guaranteed to be consistent with the input, where previous methods have no such guarantee.

cs.GR↗

Incremental Matrix Assembly for Preplanned Dynamic Tetrahedral Meshes

Local topology edits require updating assembled operators even when most of a tetrahedral mesh remains unchanged. We implement and evaluate incremental matrix assembly for preplanned edits, assuming a fixed pool of candidate elements with precomputable contributions. A preallocated sparse layout allows operator values to be updated from edited elements without rebuilding the sparsity structure. We compare full reassembly, local recomputation, and persistent incremental updates using a shared layout. Across tested meshes with up to 460k tetrahedra, incremental assembly achieves 6-27 times assembly speedups over a vectorized full-reassembly baseline for a proxy graph Laplacian and 2-17 times for linear elasticity. Frame times decrease by 1-3% when the linear solve dominates and by 18-20% in the tested synthetic dynamic elasticity loop with shorter solves. Local Recompute often achieves comparable frame times. For fixed time-step coefficients, directly maintaining the combined elasticity system also avoids a global matrix-combination pass. The Laplacian matches full reconstruction exactly in every tested frame; elasticity differences remain at rounding scale, reaching a maximum entry difference of 4.2 times 10^(-16) over 5,000 frames.

cs.GR↗

Segmenting Geometric Textures of 3D Meshes

Recognizing geometric texture patterns on 3D models is critical for applications such as heritage protection, model editing, and 3D printing. Existing 3D model segmentation methods predominantly focus on identifying high-level semantic parts, thereby failing to capture middle-level textures. While recent Vision-Language Models can perform image segmentation with appropriate text prompts, they still struggle to recognize 3D geometric textures. In this work, we present a method that segments 3D mesh models based on middle-level geometric textures. We further introduce a new geometric texture segmentation benchmark and fine-tune SAM3 accordingly. The segmented texture regions are then clustered based on DINO feature similarity and lifted from 2D to 3D. Additionally, we develop an interactive tool to enable both 3D model editing based on our segmentation results and model retrieval from the dataset based on geometric and texture similarity. Experiments demonstrate that our method significantly outperforms existing methods in segmenting geometric textures and can facilitate various applications.

cs.GR↗