A Mesh Reconstruction Method and System Based on 3D Gaussian Scatter Points

By employing a mesh reconstruction method based on 3D Gaussian scatter points and utilizing confidence filtering and iterative growth strategies, the problem of mesh fragmentation caused by noise and outliers in 3D Gaussian sputtering data reconstruction is solved, generating a high-quality triangular mesh model suitable for applications such as 3D modeling and virtual reality.

CN122312960APending Publication Date: 2026-06-30WUYI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
WUYI UNIV
Filing Date
2026-03-18
Publication Date
2026-06-30

AI Technical Summary

Technical Problem

Existing methods, when directly performing global triangulation on 3D Gaussian sputtering data, are prone to resulting in fragmented and low-quality reconstructed meshes due to noise and outliers, making it difficult to generate visually consistent and geometrically coherent high-quality mesh models.

Method used

A mesh reconstruction method based on three-dimensional Gaussian scatter points is adopted. This method involves constructing an initial triangular mesh, calculating the comprehensive confidence score, selecting high-confidence seed patches, performing iterative region growth, and then performing topology correction and smoothing. The mesh is reconstructed using geometric, appearance attributes, and photometric consistency criteria.

Benefits of technology

It generates detailed and visually consistent mesh models, reduces the dependence on the perfection of input data, improves the geometric accuracy and topological correctness of reconstruction, and outputs structured triangular mesh surfaces, which are suitable for 3D modeling, virtual reality and other fields.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122312960A_ABST
    Figure CN122312960A_ABST
Patent Text Reader

Abstract

This invention relates to the interdisciplinary field of computer vision and graphics. It provides a method and system for mesh reconstruction based on 3D Gaussian scattering. The method includes obtaining a 3D Gaussian sputtering representation of a target scene, constructing an initial triangular mesh, selecting triangular faces from the initial triangular mesh to form a high-confidence seed face set, performing iterative region growth to reconstruct the surface mesh, stopping growth when no new candidate triangular faces are accepted or the iteration termination condition is met, and performing topological correction and smoothing on the final surface mesh to output the reconstructed 3D mesh model. By employing a strategy of "high-confidence seed selection" and "multi-criteria iterative growth," the reconstruction is ensured to start from a reliable geometric region, and newly added faces are strictly constrained to maintain smooth continuity with the already reconstructed parts in terms of geometry, scale, and appearance. This avoids the spread of erroneous faces and enables the reconstruction of a mesh model with rich details and strong visual consistency.
Need to check novelty before this filing date? Find Prior Art