Point Cloud Slice Compression for Low-Latency Interactive Streaming

The patent segments point cloud data into multiple slices, where each slice contains points with similar properties (e.g., spatial location, depth range, or visual characteristics). This segmentation allows the encoder and decoder to process only relevant slices based on view direction and occlusion information, reducing the complexity of encoding/decoding while maintaining real-time processing capability for VR and self-driving services.

The patent applies different compression strategies to different slices of point cloud data based on their local characteristics. Slices that are more likely to be occluded or less visible are compressed more aggressively, while slices that are more likely to be visible are compressed less. This local quality approach reduces overall data size while minimizing the complexity increase, as the compression parameters are adapted to local slice properties rather than applying uniform compression throughout.