Adaptive Deblocking Filter for Video Block Boundaries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding techniques face issues with deblocking filtering, such as introducing boundaries inside prediction blocks and uneven filtering effects across horizontal and vertical boundaries, leading to suboptimal image fidelity and bitrate efficiency.
Innovation Solution
The proposed solution involves filtering the overlap and non-overlap areas of a block using samples from neighboring blocks and the current block, with filter strengths that decay linearly or non-linearly with distance from the boundary, allowing for smooth transitions and adaptive filtering based on block types and quantization parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If deblocking filtering is applied to reduce boundaries between blocks, then blocking artifacts are reduced, but natural structures may be removed and image fidelity deteriorates
Solution Approach 1:
The patent applies different filtering strengths to different boundary types: strong filtering for transform block boundaries, weak filtering for prediction block boundaries, and no filtering for intra-coded block boundaries. This localized differentiation preserves natural structures while removing blocking artifacts in appropriate regions.
Solution Approach 2:
The filtering strength is dynamically adjusted based on boundary type and coding mode. The system transitions between strong, weak, and no filtering depending on the specific boundary characteristics, enabling adaptive response to local image content requirements.
2Object-affected harmful factors
If deblocking filtering is applied to both horizontal and vertical boundaries, then boundary artifacts are reduced, but filtering complexity increases
Solution Approach 1:
The patent segments the filtering process into distinct stages: transform block boundary filtering followed by prediction block boundary filtering. Each stage handles specific boundary types with appropriate filtering strength, avoiding redundant filtering operations and reducing overall complexity.
Solution Approach 2:
The patent applies filtering selectively rather than uniformly across all boundaries. By applying strong filtering only where needed (transform block boundaries) and weak or no filtering elsewhere (prediction block boundaries), the system achieves effective artifact reduction without excessive computational cost.
3Ease of manufacture
If uniform filtering strength is applied to all block boundaries, then implementation is simplified, but uneven filtering effects occur across different boundary types
Solution Approach 1:
The patent implements location-dependent filtering by assigning different filtering strengths to different boundary locations and types. Transform block boundaries receive strong filtering, prediction block boundaries receive weak filtering, and intra-coded boundaries receive no filtering, achieving uniformity appropriate to each local context.
4Object-affected harmful factors
If strong filtering is applied to remove blocking artifacts, then boundary visibility is reduced, but natural structures and image details are lost
Solution Approach 1:
The patent applies strong filtering only to transform block boundaries where blocking artifacts are most prominent, while applying weak or no filtering to prediction block boundaries and intra-coded boundaries where natural structures are more likely to occur. This preserves image details in regions where they are critical.
Solution Approach 2:
The patent converts the potential harm of over-filtering into benefit by using boundary type classification to guide filtering strength. The filtering process itself becomes adaptive, using the identified boundary characteristics to determine appropriate filtering levels, thereby preventing detail loss while maintaining artifact removal effectiveness.
Data Source
AI summary
A method performed by a device for filtering a current block adjacent to an upper block and a left block is provided. The method includes (1) filtering a first sample located in a top-left area of the current block using (i) at least one sample from the upper block above the top-left area, (ii) at least one sample from the left block, and (iii) at least one unfiltered sample value for the first sample; (2) filtering a second sample located in a top-right area of the current block using (i) at least one sample from the upper block and (ii) at least one unfiltered sample value for the second sample; and (3) filtering a third sample located in a bottom-left area of the current block using (i) at least one sample from the left block and (ii) at least one unfiltered sample value for the third sample.