Point Cloud Coding With Timestamp-Ordered Output for Display Alignment

The patent changes the output order determination from a static reference-count-based approach to a dynamic timestamp-based approach. By using timestamps to track the temporal order of PC samples and outputting samples based on their timestamps rather than completing all references first, the system achieves both reliable reference handling and eliminates latency by aligning output with display timing.

The patent extends the prediction direction determination from a single dimension (parent node occupancy only) to multiple dimensions by incorporating sibling node occupancy information and planar information. This multi-dimensional approach allows the system to consider not only the parent node's occupancy status but also the occupancy patterns of sibling nodes and the planar distribution characteristics, significantly improving prediction efficiency while maintaining manageable complexity through structured information processing.

The patent segments the prediction process into distinct stages: first determining prediction directions based on parent node occupancy, then refining predictions using sibling node occupancy and planar information only when necessary. This segmented approach allows the system to use multiple reference nodes and their associated information (planar and geometry data) to improve prediction efficiency, while managing coding complexity by applying these enhanced prediction methods selectively rather than universally.