Video Processing Candidate List Construction Using Multi-Threshold Filtering

The candidate list construction process is segmented into multiple stages, each with its own threshold. The first threshold filters candidates in an initial stage, and a second threshold further refines the list in a subsequent stage. This segmentation allows the system to achieve better coding efficiency through multi-stage filtering while keeping each individual stage relatively simple.

The system dynamically adjusts threshold values based on different candidate types and processing stages. Different thresholds are applied to different groups of motion vector prediction candidates, allowing the system to optimize the balance between complexity and efficiency adaptively rather than using a fixed single threshold throughout.

Multiple reordering processes are performed in advance during the encoding stage to construct an optimized MVP candidate list. By performing these reordering operations preliminarily before actual video block processing, the system achieves better coding efficiency while the time cost is incurred during setup rather than during repeated processing operations.

The system changes the ordering parameters and criteria through multiple passes, applying different sorting rules in different stages. This allows the MVP candidate list to be optimized from multiple perspectives, improving coding efficiency by finding better candidate arrangements without requiring excessive computational resources in any single pass.