Adaptive Video Coding Filtering With Scaled Block Refinement

The patent applies adaptive filtering where filter parameters are dynamically adjusted based on local picture characteristics such as gradient calculations. The filtering strength and mode are adapted per block rather than using fixed pre-defined filters, allowing the system to optimize filtering accuracy for different regions while maintaining manageable complexity through localized adaptation.

The patent changes filtering parameters based on picture content analysis. Gradient calculations and local variance measurements are used to determine optimal filter strength and type for each block. This parameter adaptation resolves the contradiction by making filtering accuracy dependent on actual picture characteristics rather than using uniform pre-defined filters.

The patent applies filtering selectively based on local picture characteristics. Gradient calculations identify regions where filtering is most beneficial, and filtering is applied with appropriate strength only in those regions. This localized approach improves visual quality where needed while avoiding unnecessary filtering in already clean regions, thus controlling bitrate increase.

The patent applies filtering at varying strengths rather than uniformly maximum filtering. By using gradient-based adaptation to determine appropriate filter strength for each block, the system applies just enough filtering to achieve visual quality improvement without the excessive filtering that would unnecessarily increase bitrate.

The patent divides the picture into blocks and performs gradient calculations and adaptive filtering on each block independently. This segmentation allows the computationally intensive gradient-based adaptation to be distributed across multiple small regions rather than applied to the entire picture, reducing overall computational power requirements while maintaining filtering precision.

The patent performs full gradient-based adaptive filtering only where needed rather than uniformly across all blocks. By calculating gradients and applying adaptive filtering selectively based on local characteristics, the system achieves high filtering precision in regions that need it while avoiding unnecessary computational overhead in regions where simple filtering suffices.