Point Cloud Coding Prediction Method for Attribute Information

The patent applies preliminary action by performing prediction of attribute information for current nodes based on reference nodes before the actual coding process. The prediction step is executed in advance to prepare compressed data, leveraging temporal and spatial redundancy to improve coding efficiency without adding complex real-time processing during transmission.

The patent introduces prediction as an intermediary mechanism between the raw point cloud data and the final compressed bitstream. This intermediary prediction step processes attribute information using reference nodes (spatial and temporal references) to generate predicted values that reduce redundancy, thereby improving coding efficiency while maintaining a relatively simple overall coding structure.

The patent applies segmentation by dividing the prediction process into distinct components: identifying reference nodes (spatial references from neighboring nodes and temporal references from previous frames), determining prediction relationships between reference and current nodes, and generating predicted attribute values. This segmentation simplifies the overall complexity by handling each aspect separately in a systematic manner.

The patent utilizes parameter changes by adapting prediction methods based on different reference node types (spatial vs. temporal references) and their corresponding relationships. The system changes prediction parameters and approaches depending on whether spatial or temporal redundancy is leveraged, allowing efficient coding while managing complexity through parameter adaptation rather than a single complex unified approach.