GPM Motion Refinement for Video Coding Efficiency

The patent applies preliminary action by performing motion refinement on merge candidates before they are used to generate motion data for GPM coded blocks. The refinement process is conducted in advance during the candidate generation stage, so that when motion data is finally needed for coding, the refined motion information is already ready. This resolves the contradiction by preparing high-quality motion data beforehand, improving coding performance without adding complexity during the actual coding process.

The patent segments the motion estimation process into distinct stages: first generating regular merge candidates, then refining them separately, and finally using the refined motion data for GPM coding. This segmentation allows the refinement operation to be isolated as a separate processing step, improving the quality of motion data for GPM blocks while maintaining a clear and manageable workflow that doesn't unnecessarily complicate the overall coding process.

The patent performs motion refinement preliminarily during the encoding process by refining merge candidates before they are used for GPM block coding. This preliminary refinement ensures that high-accuracy motion information is prepared in advance, so that when decoding occurs, the motion data is already optimized, reducing the need for time-consuming refinement operations during the actual decoding process.

The patent uses the refined motion information from merge candidates as a template or copy that guides the motion estimation for GPM blocks. Instead of performing full refinement operations during decoding, the system uses the pre-refined motion data as a basis, effectively copying the refined characteristics to accelerate the decoding process while maintaining high prediction accuracy.