Point Cloud Context Modeling Reducing Memory Usage

The patent extracts and utilizes child-node-level occupancy information from previously coded child nodes as additional context for encoding the current parent node. This selective extraction of relevant information from a different level (child nodes instead of only parent nodes) enriches the context without proportionally increasing memory usage, as the child-node contexts are naturally available from the encoding process.

The patent transitions from traditional parent-node-level context modeling to a multi-dimensional approach that incorporates both parent-node-level and child-node-level contexts. By adding the child-node dimension to the context modeling, the system expands the receptive field and improves compression performance without linearly increasing memory consumption, since child-node information is already present in the encoding hierarchy.

The patent merges parent-node-level contexts and child-node-level contexts into a unified context modeling framework. This combination allows the system to leverage both the hierarchical parent-node structure and the detailed child-node occupancy information, achieving improved compression performance while maintaining efficient memory usage through shared context structures and selective context utilization.

The patent creates a universal context modeling approach that serves multiple functions: it maintains the traditional parent-node-level context for hierarchical structure while simultaneously incorporating child-node-level context for enhanced detail. This multi-functional context system improves compression performance across different scenarios without proportionally increasing memory requirements, as the same context structures serve multiple encoding purposes.