Flexible Deblocking Filtering for Video Subblocks

The patent applies different filtering strengths to different regions of the image based on local characteristics. Specifically, it identifies edge regions and applies stronger filtering only where blocking artifacts are most prominent, while using milder filtering in smooth regions. This localized approach maintains image quality improvement while reducing overall computational complexity compared to uniform strong filtering across the entire image.

The filtering strength is dynamically adjusted based on local image characteristics such as edge detection results and variance measurements. The algorithm adaptively determines the appropriate filtering intensity for each region, transitioning between different filtering modes (strong, medium, weak) based on real-time analysis of pixel variations and edge presence, thereby optimizing the balance between artifact removal and processing efficiency.

The patent implements region-adaptive filtering where the filtering strength varies according to local image characteristics. Edge regions detected through gradient analysis receive moderate filtering to preserve sharpness, while smooth uniform regions receive stronger filtering to effectively remove blocking artifacts. This spatially varying approach ensures that artifact reduction is achieved without uniformly compromising image sharpness and detail across the entire image.

The filtering operation applies asymmetric treatment to pixels on opposite sides of block boundaries. Instead of symmetric filtering that would blur edges equally on both sides, the algorithm detects edge orientations and applies filtering preferentially perpendicular to edges while preserving parallel edge details. This asymmetric filtering strategy maintains edge sharpness while still reducing blocking artifacts in non-edge regions.