Methods, encoders, decoders, and programs for encoding and decoding 3D point clouds.
The use of an octree structure with lexicographical ordering and a tracking table for 3D point clouds addresses inefficiencies in existing methods, enhancing memory usage and encoding/decoding efficiency by optimizing the search for adjacent nodes, thus improving compression performance and reducing latency.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- BEIJING XIAOMI MOBILE SOFTWARE CO LTD
- Filing Date
- 2022-08-11
- Publication Date
- 2026-06-02
AI Technical Summary
Existing point cloud compression methods, such as V-PCC and G-PCC, face inefficiencies in memory usage and algorithmic complexity when encoding and decoding 3D point clouds, particularly for sparse and dynamic AR/VR point clouds, due to the need to search for adjacent nodes and child nodes, which affects compression performance and latency.
The method employs an octree structure with lexicographical ordering and a tracking table to efficiently encode and decode 3D point clouds by determining adjacency patterns and entropy encoding occupancy information, reducing memory usage and complexity through raster scan ordering and a subset of neighboring nodes.
This approach minimizes memory consumption and enhances encoding/decoding efficiency by optimizing the search for adjacent nodes, thereby improving compression performance and reducing latency without sacrificing encoding characteristics.
Smart Images
Figure 0007869402000003 
Figure 0007869402000004 
Figure 0007869402000005