Cross-Component Prediction Using Pre-Processed Color Components

The patent applies preliminary action by performing pre-processing on color components before cross-component prediction. Specifically, statistical characteristics of color components are calculated and used to adjust prediction weights, ensuring that the prediction model accounts for differences in variance and correlation among components. This advance preparation improves prediction efficiency without significantly increasing overall system complexity.

The patent changes parameters by dynamically adjusting prediction weights based on statistical characteristics. The weight calculation incorporates variance and correlation coefficients of different color components, allowing the prediction model to adapt to varying statistical properties. This parameter adjustment resolves the contradiction by improving accuracy through data-driven weight optimization rather than fixed weights.

The patent applies local quality by calculating statistical characteristics locally for each prediction unit rather than globally. The variance and correlation coefficients are computed for specific color components within the current block and neighboring blocks, allowing the prediction to adapt to local statistical properties. This localized approach improves accuracy where needed while limiting computational overhead to relevant regions only.

The patent implements partial action by selectively applying statistical analysis to only the necessary color components and blocks. Rather than computing all possible statistical measures for all components, the method focuses on calculating variance and correlation for the specific components involved in cross-component prediction, achieving sufficient accuracy with reduced computational effort.