Affine Motion Compensation with Adaptive Vector Resolution

The patent implements dynamic motion vector resolution by allowing the resolution of motion vectors to vary adaptively across different regions and blocks in the video sequence. The sps_amvr_enabled_flag and related parameters enable the system to dynamically select between different resolution levels (e.g., 1/4 pel, 1/2 pel, 1 pel) based on local motion characteristics, thereby optimizing coding efficiency while managing complexity through conditional application rather than uniform high-resolution processing throughout the entire sequence

The patent applies local quality by enabling AMVR on a block-by-block or region-by-region basis using flags such as inter_affine_flag and amvr_enabled_flag. Different blocks within the same picture can use different motion vector resolutions according to their specific motion complexity requirements. This allows high-resolution motion vectors to be applied only where necessary (e.g., regions with complex motion) while using lower resolution elsewhere, thus improving overall coding efficiency without uniformly increasing processing complexity across the entire video sequence

The patent employs preliminary action by defining a hierarchical flag structure where high-level flags (sps_amvr_enabled_flag, sps_affine_enabled_flag) are parsed first at the sequence or picture level to establish whether AMVR and affine motion compensation are enabled. Based on these preliminary decisions, lower-level flags (amvr_enabled_flag, inter_affine_flag, affine_amvr_enabled_flag) are conditionally parsed at the block level. This preliminary action reduces the average number of flags that need to be parsed by eliminating unnecessary parsing when features are not enabled, thereby enhancing compression efficiency while managing parsing complexity through conditional execution

The patent segments the flag parsing process into distinct hierarchical levels: sequence-level flags (sps_amvr_enabled_flag), picture-level flags (amvr_enabled_flag), and block-level flags (inter_affine_flag, affine_amvr_enabled_flag). This segmentation allows the decoder to parse only the relevant flags for each specific context rather than processing all possible flags uniformly. The segmented approach enables efficient video compression by applying complex features like affine motion compensation with AMVR only where needed, while reducing parsing complexity through conditional flag processing at each hierarchical level