Mesh Denoising Based on Normal Voting Tensor and Binary Optimization.

Area of Science:

• Computer Graphics
• Computational Geometry
• Digital Signal Processing

Background:

• 3D mesh data often suffers from noise, degrading its visual quality and utility.
• Traditional denoising methods like averaging can blur sharp features and create undesirable artifacts.
• Existing algorithms struggle to balance surface smoothing with feature preservation.

Purpose of the Study:

• To develop an advanced mesh denoising algorithm that overcomes limitations of current methods.
• To enhance the preservation of sharp geometric features during the denoising process.
• To achieve smoother surfaces, particularly in complex regions like umbilical points.

Main Methods:

• A two-stage mesh denoising approach utilizing an element-based normal voting tensor.
• Incorporation of binary optimization and a local binary neighborhood concept.
• Stochastic analysis of noise based on average edge length for noise characterization.

Main Results:

• The proposed algorithm effectively retains sharp features, unlike traditional averaging methods.
• Smoother umbilical regions are produced, indicating improved handling of complex geometry.
• Quantitative results show superior performance compared to state-of-the-art smoothing approaches.

Conclusions:

• The novel two-stage algorithm offers significant improvements in mesh denoising.
• It provides a robust solution for preserving geometric details while achieving surface smoothness.
• This method represents a advancement in 3D mesh processing and noise reduction techniques.