Surface Reconstruction from Point Clouds via Grid-based Intersection Prediction

TL;DR

This work tackles the challenge of reconstructing surfaces from point clouds, particularly open surfaces, by directly predicting the intersection points between cube edges and implicit surfaces rather than relying on UDF values. It introduces two symmetry-driven modules—the Relative Sign Module and the Intersection Module—to ensure surface continuity and reduce artefacts, and reconstructs meshes via a Marching Cubes template. The method achieves state-of-the-art performance across ShapeNet, MGN, and ScanNet, demonstrating strong results on watertight, open, and real-scene data, with notable improvements in Chamfer Distance and normal consistency. The approach offers practical impact for high-fidelity surface reconstruction in vision and graphics, with future work aimed at handling edge cases near open-surface boundaries more effectively.

Abstract

Surface reconstruction from point clouds is a crucial task in the fields of computer vision and computer graphics. SDF-based methods excel at reconstructing smooth meshes with minimal error and artefacts but struggle with representing open surfaces. On the other hand, UDF-based methods can effectively represent open surfaces but often introduce noise, leading to artefacts in the mesh. In this work, we propose a novel approach that directly predicts the intersection points between line segment of point pairs and implicit surfaces. To achieve it, we propose two modules named Relative Intersection Module and Sign Module respectively with the feature of point pair as input. To preserve the continuity of the surface, we also integrate symmetry into the two modules, which means the position of predicted intersection will not change even if the input order of the point pair changes. This method not only preserves the ability to represent open surfaces but also eliminates most artefacts on the mesh. Our approach demonstrates state-of-the-art performance on three datasets: ShapeNet, MGN, and ScanNet. The code will be made available upon acceptance.

Figures (13)

• Figure 1: The left column is shape visualization when we keep the vertices sign at the corner of cube correct and give the intersection a random disturbing. The right column is the GT shape.
• Figure 2: This figure indicates the two parts in reconstructing mesh, the sign of vertices of 8 corners in a cube in (a) and the intersection position in (b).
• Figure 3: The pipeline of our method. First, we predict the relative sign of the vertices at the corners of the cube. Then, we predict the intersection between the line segment of the point pair and the implicit surface. Finally, we construct the mesh using the template in Marching Cubes.
• Figure 4: The network architecture of our method, where $M$ is the number of point pair, $N$ is the number of points in a point cloud.
• Figure 5: The symmetry ensures the continuity of surface.