Deblocking Filter Gradient Assessment and PU Boundary Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional deblocking filters in video coding apply unnecessary filtering and do not accurately assess edge gradients, leading to suboptimal performance and edge sharpness preservation.
Innovation Solution
A method that determines the existence of non-zero transform coefficients for PU-only boundaries and uses a combination of gradients from multiple pixel lines to decide on filter ON/OFF and strength, with conditional application of weak filters to preserve edge sharpness, and adjusts filter strength based on specific gradient thresholds and quantization parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional deblocking filter is applied to all boundaries, then coding artifacts are reduced, but unnecessary filtering is applied and edge sharpness is compromised
Solution Approach 1:
The patent applies different filtering strategies to different boundary types (PU-only boundaries vs. TU boundaries). PU-only boundaries receive weaker or no filtering while TU boundaries receive standard deblocking filtering, allowing localized optimization that reduces unnecessary filtering while maintaining artifact reduction where needed.
Solution Approach 2:
The patent dynamically adjusts filter strength based on boundary characteristics and transform coefficient properties. The filter strength is not fixed but adapts based on the specific conditions of each boundary, enabling flexible control that prevents unnecessary filtering while maintaining effectiveness.
2Reliability
If filter strength is increased to reduce artifacts, then coding performance improves, but edge sharpness is lost
Solution Approach 1:
The patent applies different filter strengths to different regions and boundary types. Strong filtering is applied only where necessary (TU boundaries with transform coefficients), while weak or no filtering is applied to PU-only boundaries, preserving edge sharpness in regions where it matters most.
Solution Approach 2:
The patent changes the filter strength parameter dynamically based on boundary characteristics. By adjusting the filter strength parameter according to the specific conditions of each boundary, the system optimizes the balance between artifact reduction and edge sharpness preservation.
3Reliability
If deblocking filter is applied to PU-only boundaries, then boundary artifacts are reduced, but computational complexity increases unnecessarily
Solution Approach 1:
The patent extracts and separates PU-only boundaries from TU boundaries, applying different processing rules to each. By taking out PU-only boundaries and applying weaker or no filtering, the system reduces unnecessary computational complexity while maintaining effectiveness where needed.
Solution Approach 2:
The patent segments the boundary processing into different categories (PU-only vs. TU boundaries) and applies appropriate filtering strategies to each segment. This segmentation allows computational resources to be allocated efficiently, avoiding unnecessary processing on PU-only boundaries.
Data Source
AI summary
A method and apparatus for improved deblocking filter are disclosed. In one embodiment according to the present invention, determining the non-zero transform coefficients existence for a PU-only boundary is skipped and therefore the deblocking filter for PU-only boundary is eliminated if the condition of boundary strength equal to 1 is not fulfilled to avoid unnecessary deblocking. In another embodiment of the present invention, three gradients are determined for two pixel lines and filter ON/OFF decision is made based on the three gradient values of the two pixel lines. One aspect of the invention is related to filter strength selection between strong and weak filters. According to one embodiment, the filter strength selection is based on various gradient values of a pixel line across the block boundary. Another aspect of the invention is related to the weak filter.