Point Cloud Attribute Prediction Modes for Efficient Compression

The patent applies preliminary action by performing temporal prediction using motion compensation before the main RAHT encoding process. The encoder predicts attribute values at current time based on previously encoded attributes at previous times, and encodes only the prediction residuals. This preliminary temporal prediction step removes redundancy before the spatial RAHT transform, thereby improving overall encoding efficiency while maintaining the RAHT framework.

The patent implements partial action by providing multiple prediction mode flags (intra prediction mode flag and inter prediction mode flag) that can be selectively used depending on the encoding scenario. The encoder can choose to use intra prediction, inter prediction, or both, with conditional encoding of their respective flags. This provides flexibility while controlling complexity by only fully implementing the prediction modes that are most beneficial for different types of point cloud data.