Video Decoder DMVR Flag Control for Efficient Bi-Prediction

The patent applies different quality enhancement techniques (DMVR and BDOF) to different regions or blocks of the video picture based on local characteristics. By analyzing motion complexity and other local features, the encoder selectively applies these computationally intensive techniques only where they provide significant quality improvement, rather than uniformly across the entire picture. This local differentiation resolves the contradiction by concentrating computational resources on regions that benefit most from refinement.

The patent implements a selective application approach where DMVR and BDOF are applied partially to only certain picture blocks that meet specific criteria, rather than excessively applying them to all blocks. The encoder uses flags and conditional logic to determine which blocks warrant the additional computational effort, performing the refinement action only to the extent necessary for achieving quality improvement while avoiding unnecessary computational overhead.

The patent introduces syntax elements at the picture parameter set (PPS) level that allow different regions or blocks within a picture to have different DMVR and BDOF application settings. This local quality control enables the encoder to maintain high picture quality in complex motion regions while using simpler coding methods in uniform regions, thereby improving overall coding efficiency without sacrificing quality where it matters most.

The patent makes the application of DMVR and BDOF dynamic rather than static by introducing flags that can be set on a per-picture or per-block basis. These dynamic flags allow the encoding strategy to adapt to the specific content characteristics of each picture or region, enabling the system to switch between high-quality refinement and efficient coding modes as needed, thus resolving the contradiction between quality and efficiency.

The patent implements a nested syntax structure where DMVR and BDOF flags are defined at multiple hierarchical levels: the sequence parameter set (SPS) level for global defaults, and the picture parameter set (PPS) level for picture-specific or slice-specific overrides. This nested structure allows the system to maintain simple global configuration when uniform application is desired, while enabling fine-grained selective application when needed, thus resolving the contradiction between operational simplicity and adaptability.

The patent designs the syntax structure to serve multiple functions: the SPS-level flags provide universal application across all pictures for simplicity, while the PPS-level flags provide specialized control for specific pictures or slices when optimization is needed. This multi-functional syntax design allows the same basic mechanism to operate at different granularities, resolving the contradiction between ease of operation and selective application capability.