GPF: GMM-Inspired Feature-Preserving Point Set Filtering.

Area of Science:

• Computer Science
• Geometry Processing
• Computer Vision

Background:

• Point set filtering is crucial for reconstructing clean 3D data from noisy inputs.
• Existing methods often fail to preserve fine-scale geometric features or sharp edges.
• Balancing noise removal and feature preservation remains a significant challenge.

Purpose of the Study:

• To develop a robust and effective point set filtering method.
• To improve the preservation of geometric features during noise reduction.
• To outperform current state-of-the-art filtering techniques.

Main Methods:

• A novel approach inspired by the Gaussian Mixture Model (GMM).
• Utilizes noisy point sets and their filtered normals as input.
• Reconstructs a high-quality, noise-free, and feature-preserving point set.

Main Results:

• The proposed method successfully filters noise from point sets.
• It demonstrates superior preservation of geometric features, including fine details.
• Experimental results show significant improvements over existing methods in filtering quality and accuracy.

Conclusions:

• The GMM-inspired method offers a robust solution for feature-preserving point set filtering.
• This approach enhances the quality and accuracy of 3D geometry reconstruction.
• It addresses limitations of current methods in handling complex geometric features.