Graph-Based Feature-Preserving Mesh Normal Filtering.

Area of Science:

• Computer Graphics
• Computational Geometry
• Image Processing

Background:

• Accurate mesh denoising is crucial for preserving geometric details.
• Distinguishing features from noise remains a significant challenge in mesh processing.

Purpose of the Study:

• To propose a graph-based feature-preserving mesh normal filtering scheme.
• To enhance mesh denoising by robustly identifying and preserving geometric features.

Main Methods:

• Representing mesh faces as patches and modeling them as weighted graphs for feature detection.
• Employing a graph-cut and iterative normalized cut algorithm for feature-aware region segmentation.
• Utilizing feature-aware guided normal filtering for robust denoising.

Main Results:

• The proposed scheme successfully distinguishes geometric features from noise in noisy meshes.
• Experimental results demonstrate superior performance compared to state-of-the-art methods.
• Both objective and subjective evaluations confirm the effectiveness of the feature-preserving approach.

Conclusions:

• The developed graph-based filtering scheme offers robust and feature-preserving mesh denoising.
• This method significantly advances the state-of-the-art in mesh normal filtering.
• The approach is effective for both synthetic and real-world scanned models.