Adaptive Reference Sample Filtering in Weighted Intra Prediction

The patent dynamically adjusts the filtering strength or applicability of reference sample filtering based on the intra prediction mode being used. When weight-based intra prediction is detected, the filtering is adapted or disabled to preserve prediction accuracy, while for traditional intra prediction modes, filtering is applied to achieve smoothness. This dynamic adaptation resolves the contradiction by making the filtering process conditional rather than static.

The patent changes the filtering parameters (such as filtering strength, filter type, or application condition) based on the prediction mode. By modifying these parameters adaptively, the system achieves both smoothness when appropriate and accuracy when needed, resolving the contradiction between these two opposing requirements.

The patent applies weight-based intra prediction selectively to specific regions or blocks where it provides the most benefit, rather than uniformly across the entire image. By localizing the complex processing to only where needed, the system achieves improved prediction accuracy in critical areas while minimizing overall processing complexity.

The patent implements a simplified or partial version of weight-based intra prediction that achieves sufficient accuracy without the full computational overhead. This may involve using reduced sets of reference samples, simplified weighting calculations, or selective application to certain prediction modes, thereby balancing accuracy improvement with complexity reduction.