Merge Mode Signaling for Low-Overhead Video Motion Prediction

The patent implements dynamic enablement of merge modes where the encoder and decoder can adaptively select which merge modes are available based on video content characteristics and complexity constraints. The system dynamically adjusts the set of enabled merge modes rather than using a fixed configuration, allowing optimization of coding efficiency while managing device complexity through conditional activation of advanced modes.

The patent changes the parameter of merge mode enablement status from a fixed state to a variable state that can be adjusted based on coding conditions. By introducing parameters that control which merge modes are enabled (such as flags indicating availability of CIIP merge mode, MMVD, affine motion compensation), the system can optimize between coding efficiency and complexity by selectively enabling modes based on content requirements.

The patent extracts only the essential merge mode flags that need to be explicitly signaled, rather than signaling all possible merge mode parameters. By identifying and signaling only the critical flags (such as ciip_merge_mode_flag, mmvd_merge_mode_flag, affine_merge_mode_flag) while inferring the status of other modes, the system reduces signaling overhead while maintaining sufficient decoding accuracy for reconstructing the video data.

The patent implements self-service through inference mechanisms where the decoder automatically determines the enablement status of merge modes that are not explicitly signaled, based on the signaled flags and predefined rules. This self-service approach allows the system to maintain accurate decoding by inferring missing information from available data, thereby reducing bit rate without sacrificing decoding precision.

The patent applies dynamic selection of prediction methods where advanced techniques like CIIP (combined inter and intra prediction) and affine motion compensation are conditionally enabled based on content characteristics and complexity constraints. The system dynamically switches between simple and advanced prediction methods rather than always using the most complex algorithms, optimizing the balance between video quality and computational complexity through adaptive method selection.

The patent applies local quality by using advanced prediction methods selectively in specific regions or blocks where they provide the most benefit, rather than uniformly applying complex algorithms across the entire video frame. By identifying regions that require higher prediction accuracy (such as areas with complex motion patterns) and applying advanced methods like affine motion compensation only there, the system improves video quality in critical areas while reducing overall computational complexity.