Mesh Denoising via Adaptive Consistent Neighborhood

Area of Science:

• Computer Graphics
• Geometric Modeling
• Image Processing

Background:

• Mesh denoising is crucial for 3D data processing.
• Existing methods struggle with preserving fine geometric details and handling complex shapes.
• Guided normal filtering offers potential but requires accurate neighborhood information.

Purpose of the Study:

• To develop an advanced mesh denoising technique.
• To improve the accuracy of guided normal filtering by creating adaptive neighborhoods.
• To enhance feature preservation and robustness for complex mesh structures.

Main Methods:

• A two-stage scheme for constructing adaptive consistent neighborhoods.
• A novel consistency measurement for initial neighborhood selection (patch-shift manner).
• An iterative graph-cut based scheme to refine neighborhoods, removing geometric features.
• Guided normal filtering utilizing these adaptive neighborhoods.
• Vertex updating for final mesh refinement.

Main Results:

• The proposed method effectively removes noise from 3D meshes.
• Geometric features are well-preserved, even on complex surfaces.
• The adaptive consistent neighborhoods lead to a more accurate guide normal field.
• Quantitative and visual experiments demonstrate superior performance compared to existing methods.

Conclusions:

• The novel guided normal filtering with adaptive neighborhoods significantly advances mesh denoising.
• The method is robust and excels at preserving geometric details.
• This approach offers a superior solution for processing noisy 3D mesh data.