Luma-Derived Chroma Intra Prediction for Video Compression

The patent divides the picture into multiple coding tree units (CTUs) that can be further recursively split into smaller coding units (CUs) at different depths. This hierarchical segmentation allows the encoder to adaptively select the appropriate granularity for different regions, improving encoding coverage for high-resolution video while avoiding unnecessary complexity in uniform areas.

The patent implements dynamic adaptation of coding unit sizes and prediction modes based on local image characteristics. The encoder can adjust the depth of hierarchical splitting and select different intra prediction modes (including directional modes derived from luminance components) for chrominance components, optimizing compression efficiency for each specific region rather than using a fixed approach throughout the entire picture.

The patent employs a hierarchical structure where the picture is divided into CTUs that can be recursively split into smaller CUs. This creates multiple levels of segmentation that allow the encoder to process different regions at appropriate granularities, improving compression efficiency without requiring complete redesign of the encoding process.

The patent changes the parameter of coding unit size from fixed to variable by introducing hierarchical splitting with different depths. The encoder can adjust the depth parameter to control the size of coding units, and can also change prediction modes (including deriving chroma intra prediction modes from luma modes) based on local characteristics, thereby improving compression efficiency while maintaining reasonable encoding complexity.

The patent makes the intra prediction mode determination process universal by deriving chrominance component prediction modes from luminance component prediction modes. Instead of independently determining prediction modes for both luma and chroma, the system uses the luma prediction mode as a base and applies it to chroma components, reducing encoding complexity while maintaining adequate prediction accuracy through adaptive adjustments when needed.

The patent copies the intra prediction mode determined for the luminance component and applies it to the chrominance component. This copying approach reduces the number of independent mode decisions required, lowering encoding complexity and bit rate while still achieving acceptable prediction accuracy for chrominance, which is less sensitive to prediction errors than luminance.