Point Cloud Occupancy Coding With Reduced Neighbor Contexts

The patent segments the volumetric space into a tree structure (octree) where each node represents a sub-volume. This segmentation allows the coding process to operate on smaller, more manageable units rather than the entire point cloud at once, reducing the context management burden while maintaining compression efficiency through localized context-adaptive binary entropy coding of occupancy patterns.

The patent applies different context models locally to different regions of the point cloud based on occupancy patterns and neighbor configurations. By adapting the context selection to local characteristics (occupied vs. unoccupied neighbors, position in tree structure), the system achieves high compression efficiency without requiring a single overly complex global context model.

The patent performs preliminary classification of neighbor configurations into shielded and unshielded categories before the actual coding process. By pre-determining which neighbors can shield (block the line of sight to) the current node, the system can selectively apply context reduction only where appropriate, maintaining compression performance in critical regions while reducing complexity where possible.

The patent dynamically changes the context parameters based on the shielded status and neighbor occupancy patterns. Instead of using a fixed small number of contexts, the system adjusts the effective number of contexts used at each node based on local geometric characteristics, achieving a balance between complexity and performance through parameter adaptation rather than fixed reduction.