Unified Mesh and Point Cloud Coding via Cross-Reference Prediction

The patent combines mesh and point cloud compression into a unified framework where both data types are processed together. The encoder generates both a mesh representation and a point cloud representation, and the decoder reconstructs both simultaneously, allowing them to complement each other's strengths and achieve higher coding efficiency than separate techniques could provide.

The unified compression technique serves multiple functions: it compresses both mesh and point cloud data, enables bidirectional prediction between the two representations, and provides a flexible framework that can adapt to different application requirements. This multi-functional approach resolves the contradiction by integrating complexity into a universal solution that delivers superior overall performance.

The patent implements feedback mechanisms where the reconstructed mesh is used to improve point cloud decoding and vice versa. The decoder uses the reconstructed mesh to guide point cloud reconstruction, and the reconstructed point cloud can inform mesh refinement, creating a feedback loop that progressively improves both representations and utilizes their complementary information.

The encoder performs preliminary actions by generating both mesh and point cloud representations simultaneously during the encoding process. This preliminary generation of both data types allows the decoder to have both representations available for mutual reference and improvement, ensuring that complementary information is captured and utilized from the outset rather than requiring separate processing.