Dynamic Point Cloud Compression Based on Projections, Surface Reconstruction and Video Compression

Area of Science:

• Computer Vision
• Computer Graphics
• Data Compression

Background:

• Dynamic point clouds require efficient compression methods.
• Existing methods may not optimally handle geometry and texture data simultaneously.
• Advanced surface reconstruction techniques are crucial for accurate visualization.

Purpose of the Study:

• To develop and evaluate a novel dynamic point cloud compression technique.
• To investigate the impact of different projection types and bit depths on compression efficiency and quality.
• To integrate surface reconstruction with video compression for enhanced point cloud representation.

Main Methods:

• Compression using projection mapping (cylindrical, Miller, Mercator) and varying bit depths.
• Texture map compression via Voronoi diagrams.
• Application of specialized video codecs: FFV1 for geometry, H.265/HEVC for texture.
• Point cloud reconstruction using Poisson surface reconstruction.
• Quality assessment via point-to-point and point-to-plane metrics.

Main Results:

• The proposed dynamic point cloud compression method demonstrates effectiveness.
• Cylindrical, Miller, and Mercator projections yield improved results compared to others.
• The combination of projection mapping, Voronoi diagrams, and specific video codecs achieves high-fidelity reconstruction.
• Quantitative analysis confirms the performance benefits of selected projection types.

Conclusions:

• Dynamic point cloud compression can be significantly improved using projection mapping techniques.
• The choice of projection significantly impacts the efficiency and quality of reconstructed point clouds.
• The integrated approach offers a promising solution for storing and transmitting dynamic 3D data.