Hierarchical Prior-Based Super Resolution for Point Cloud Geometry Compression.

Area of Science:

• Computer Vision
• Signal Processing
• Data Compression

Background:

• Geometry-based Point Cloud Compression (G-PCC) is essential for efficient data handling.
• Lossy G-PCC methods often introduce distortions due to simple geometry quantization like grid downsampling.

Purpose of the Study:

• To propose a novel hierarchical prior-based super-resolution method for enhancing point cloud geometry compression.
• To mitigate distortions in lossy G-PCC and improve reconstruction quality.

Main Methods:

• A content-dependent hierarchical prior is constructed at the encoder.
• This prior facilitates a coarse-to-fine super-resolution process at the decoder.
• The method's performance is evaluated using the MPEG Cat1A dataset.

Main Results:

• The proposed method achieves substantial Bjøntegaard-delta bitrate savings.
• Performance surpasses existing octree-based and trisoup-based G-PCC v14 methods.
• Experimental results confirm improved reconstruction accuracy.

Conclusions:

• Hierarchical prior-based super-resolution offers a significant advancement in point cloud geometry compression.
• The method effectively reduces distortions and improves compression efficiency over current standards.
• Open-source implementations are provided for reproducibility.