Filmsticking++: Rapid Film Sticking for Explicit Surface Reconstruction

TL;DR

This work tackles explicit surface reconstruction from unoriented, potentially low-quality point clouds by addressing the limitations of prior RVD-based filmsticking methods in handling deep cavities and complex topology. It introduces Filmsticking++, which replaces the restricted Voronoi diagram with a restricted power diagram (RPD), augments the guiding surface with internal virtual sites to accelerate convergence, and applies a smoothness-aware manifold fix to disambiguate plate-like structures. The approach achieves a watertight, manifold surface with fewer iterations, improved robustness, and better scalability across synthetic and real datasets, including challenging high-genus and irregular point distributions. Practically, Filmsticking++ reduces computational cost while preserving sharp features and geometric details, enabling more reliable explicit surface reconstructions from imperfect data, though it still faces limitations on open surfaces and very thin neck regions.

Abstract

Explicit surface reconstruction aims to generate a surface mesh that exactly interpolates a given point cloud. This requirement is crucial when the point cloud must lie non-negotiably on the final surface to preserve sharp features and fine geometric details. However, the task becomes substantially challenging with low-quality point clouds, due to inherent reconstruction ambiguities compounded by combinatorial complexity. A previous method using filmsticking technique by iteratively compute restricted Voronoi diagram to address these issues, ensures to produce a watertight manifold, setting a new benchmark as the state-of-the-art (SOTA) technique. Unfortunately, RVD-based filmsticking is inability to interpolate all points in the case of deep internal cavities, resulting in very likely is the generation of faulty topology. The cause of this issue is that RVD-based filmsticking has inherent limitations due to Euclidean distance metrics. In this paper, we extend the filmsticking technique, named Filmsticking++. Filmsticking++ reconstructing an explicit surface from points without normals. On one hand, Filmsticking++ break through the inherent limitations of Euclidean distance by employing a weighted-distance-based Restricted Power Diagram, which guarantees that all points are interpolated. On the other hand, we observe that as the guiding surface increasingly approximates the target shape, the external medial axis is gradually expelled outside the guiding surface. Building on this observation, we propose placing virtual sites inside the guiding surface to accelerate the expulsion of the external medial axis from its interior. To summarize, contrary to the SOTA method, Filmsticking++ demonstrates multiple benefits, including decreases computational cost, improved robustness and scalability.

Figures (17)

• Figure 1: The state-of-the-art (SOTA) method wang2022restricted falls short in attracting the guiding surface to adhere to the internal cavity, resulting in the need for multiple cycles of filmsticking and sculpting operations. In contrast, our only use 13 Filmsticking++ and employs just one field-based dangling removal operation, significantly reduces the number of iterations and generate a high-fidelity mesh surfaces.
• Figure 2: The pipeline of our algorithm. (a) A point cloud as input. (b) The bounding sphere as an initial guiding surface as well as a set of sampling points (typically around 1K) inside the guiding surface. (c) An iterative process until the guiding surface interpolates nearly all points. (d) Field-based dangling removal for fixing the topology. (e) A single Filmsticking++ operation to accomplish the reconstruction.
• Figure 3: By replacing RVD with RPD, we can attract point that far away from guiding surface to the guiding surface. (a) Point $a$, $b$ is near the guiding surface, thus dominating an area of guiding surface. (b) With RVD, point $a$, $b$ is embedded into the dual triangulation, but dominated region of point $c$ has no intersection with the guiding surface, the point $c$ cannot be attracted onto the guiding surface. (c) With RPD, the weight of $c$ is increased so that it can be attracted to the surface successfully. In the case of RVD, only those points near the guiding surface can dominate an area. In this paper, we extend RVD to RPD, allowing distant points to adhere to the guiding surface by increasing their weights.
• Figure 4: Virtual vertices to accelerate the evolution of guiding surface. (a) Input point cloud. (b) Sample a set of virtual sites (colored in green) inside the guiding surface, to test role of virtual points, deep internal cavities are set up with and without virtual sites, respectively. (c) Update Voronoi diagram including virtual sites. (d) One Filmsticking++, a virtual site is attracted onto the guiding surface, and update the guiding surface. (e) The virtual site (yellow) is removed, allowing the guiding surface to be re-partitioned by the neighboring sites. (f) After 9 Filmsticking++, guiding surface attracted all points in deep internal cavities with virtual sites.
• Figure 5: Smoothness manifold fix. When a site (black) dominates two separate regions of the guiding surface, resulting in RDT a non-manifold vertex. (a) Filmsticking wang2022restricted determines the dominated region of a site based on the distances from the site to its multiple disconnected dominated regions, it may cause many unwanted topological. (b) We propose a smoothness-driven principle to select one of the regions and associate the site with the selected region.