Mode-Dependent Transform Signaling for Video Encoding

The patent applies local quality by differentiating the treatment of intra prediction modes based on their specific characteristics. Instead of uniformly testing all modes with EMT and NSST, the patent identifies specific modes (angular modes vs. non-angular modes) and applies transforms selectively. This allows optimized compression for modes that benefit from EMT/NSST while avoiding unnecessary complexity for modes where default transforms suffice.

The patent implements dynamics through adaptive transform selection based on intra prediction mode characteristics. The system dynamically decides whether to apply EMT, NSST, or default transforms by evaluating the prediction mode type. This dynamic approach enables the encoder to adjust its behavior based on local characteristics of each prediction mode, balancing compression efficiency with encoding complexity.

The patent applies local quality by differentiating signaling requirements based on intra prediction mode characteristics. Transform type signaling is selectively applied only to specific modes (angular modes) that benefit from it, while default transforms are used for other modes (non-angular modes) without requiring explicit signaling. This selective approach reduces the overall bit requirements while maintaining compression efficiency for the most beneficial cases.

The patent implements partial action by applying transform type signaling only to a subset of intra prediction modes rather than all modes. Specifically, signaling is applied partially to angular modes while default transforms are used for non-angular modes. This partial application reduces the total number of bits required for signaling while still providing optimized compression where most needed.

The patent applies local quality by differentiating computational processing based on intra prediction mode characteristics. Instead of uniformly applying EMT and NSST to all modes, the patent selectively applies these computationally intensive transforms only to angular modes that benefit from them, while using simpler default transforms for non-angular modes. This reduces overall computational overhead while maintaining compression efficiency for the most beneficial cases.

The patent implements partial action by applying complex transforms (EMT and NSST) only to a subset of intra prediction modes rather than all modes. Specifically, these computationally intensive transforms are applied partially to angular modes while simpler transforms are used for non-angular modes. This partial application significantly reduces computational overhead while still providing optimized compression where most needed.