A Color- and Geometric-Feature-Based Approach for Denoising Three-Dimensional Cultural Relic Point Clouds

Area of Science:

• Computer Vision
• Digital Heritage
• Signal Processing

Background:

• 3D cultural relic acquisition often results in noisy point clouds.
• High-quality 3D models are crucial for preservation and study.
• Existing denoising methods may struggle with preserving fine details.

Purpose of the Study:

• To develop an effective point cloud denoising method for 3D cultural relics.
• To combine color and geometric features for improved noise removal.
• To enhance the realism of 3D cultural relic models.

Main Methods:

• Utilized graph signal processing (GSP) on self-similar patches of 3D point clouds.
• Constructed a graph with Markov property, incorporating 3D coordinates, normal vectors, and color features.
• Formulated denoising as a maximum a posteriori (MAP) estimation with graph Laplacian regularization (GLR) prior.
• Incorporated moderate 3D point simplification to optimize processing time.

Main Results:

• The proposed GSP approach significantly outperformed five competing methods in subjective and objective evaluations.
• Demonstrated effective noise removal from cultural relic point clouds.
• Successfully preserved fine-scale 3D features, including texture and ornamentation.
• Achieved more realistic 3D representations compared to existing techniques.

Conclusions:

• The novel GSP-based method offers robust and efficient denoising for 3D cultural relic point clouds.
• Preservation of intricate details leads to enhanced realism in digital heritage models.
• The approach provides a valuable tool for digital archaeology and cultural heritage preservation.