Video De-blocking Filter Using Local Quality Adaptation

The patent applies different filter strengths to different regions of the video data based on their characteristics. Strong filtering is applied to regions with smooth transitions and weak filtering is applied to regions with high-frequency content, such as screen content. This local differentiation allows the system to maintain coding efficiency in smooth regions while preserving high-frequency details in screen content regions, thereby reducing artefacts without sacrificing compression performance.

The patent dynamically adjusts the filter strength parameter based on the quantisation parameter and prediction mode. When the quantisation parameter indicates higher quality requirements or when intra-prediction modes are detected (which often correspond to screen content), the filter strength is reduced. This parameter adaptation enables the system to balance coding efficiency with artefact reduction by tuning the filtering intensity according to local quality requirements.

The patent identifies regions with high-frequency characteristics (such as screen content with sharp edges and fine details) and applies weaker filtering to these regions while maintaining strong filtering in smooth regions. This local quality differentiation ensures that high-frequency information is preserved where needed, preventing information loss in screen content while still benefiting from compression in other areas.

The patent modifies the filter strength parameter based on detected prediction modes and quantisation parameters. When intra-prediction modes are detected (common in screen content) or when quality parameters indicate the need for detail preservation, the filter strength is reduced to prevent high-frequency information loss. This dynamic parameter adjustment ensures that coding efficiency does not come at the cost of sacrificing important high-frequency details.

The patent applies different filtering strengths to different spatial locations and regions based on their content characteristics. Regions with smooth transitions receive strong filtering to reduce artefacts, while regions with high-frequency content (screen content, text, fine details) receive weak or no filtering to preserve high-frequency details. This local quality adaptation simultaneously improves overall video quality while maintaining important high-frequency information.

The patent dynamically adjusts the filter strength parameter based on quantisation parameters and prediction modes to balance artefact reduction with high-frequency detail preservation. When quality parameters indicate the need for detail preservation or when screen content is detected, the filter strength is reduced to avoid degrading high-frequency details. This parameter adaptation ensures that video quality improvement does not come at the cost of losing important high-frequency information.