Voxel Coordinate Prediction for Smaller AR/VR Geometry Bitstreams

The patent applies preliminary action by predicting voxel coordinates before transmission. A predictor module estimates upcoming voxel coordinates based on previous values, and only transmits correction data when predictions differ from actual values. This reduces the amount of data that needs to be transmitted while maintaining full coordinate precision at the receiver end.

The patent uses copying by transmitting predicted coordinate values instead of actual values. The receiver reconstructs the full coordinate sequence by combining received correction data with locally generated predictions, creating a copy of the original coordinate information in a compressed form.

The patent applies self-service by using the sequential structure of voxel coordinate data to its advantage. The predictor module automatically generates predictions based on the inherent redundancy in consecutive voxel coordinates, and the encoding scheme adapts to exploit this self-generated pattern without requiring external entropy models.

The patent changes the parameter representation by transitioning from direct coordinate transmission to differential encoding. Instead of transmitting absolute coordinate values, the system transmits differences between predicted and actual values, fundamentally changing how coordinate data is parameterized and encoded.

The predictor module performs preliminary action by pre-calculating expected voxel coordinates for large spatial volumes. This allows the system to handle increased voxel counts without proportionally increasing bitstream size, as most coordinates are generated locally rather than transmitted.

The patent segments the coordinate transmission into prediction components (generated locally) and correction components (transmitted). This segmentation allows the system to scale to large voxel volumes while maintaining efficient transmission, as only the correction segments need to be transmitted over the network.