Adaptive Hadamard Filtering for Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems, such as HEVC, face limitations in coding efficiency due to dependencies on neighboring blocks for filtering, small filter kernel sizes, uniform filter strength across all samples, and lack of adaptation based on coding mode and prediction mode.
Innovation Solution
The proposed solution involves adaptive Hadamard filtering of reconstructed coding units, where extrapolated samples are generated to extend the current coding unit, and different filter strengths are applied based on the position of samples within the coding unit and the prediction mode used.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Length of stationary object
If filtering is performed using existing HEVC methods, then filtering can be applied to reconstructed samples, but the filter kernel size is limited and dependency on neighboring blocks increases
Solution Approach 1:
The patent applies preliminary action by generating extrapolated samples before filtering operations. These extrapolated samples extend the current coding unit boundaries, allowing the filter to operate with a larger kernel size without requiring access to neighboring blocks. The extrapolation is performed in advance, creating virtual sample values that enable extended filtering while maintaining independence from neighboring block data.
2Adaptability or versatility
If uniform filter strength is applied to all samples, then filtering can be implemented simply, but adaptability to different coding modes and prediction modes is lost
Solution Approach 1:
The patent implements local quality by applying different filter strengths to different samples based on their position within the coding unit and the prediction mode used. Instead of a uniform filter strength, the filter adapts its strength locally for each sample or sample group, allowing optimization for specific regions and coding modes while maintaining a manageable implementation complexity through systematic classification.
Solution Approach 2:
The patent applies dynamics by making the filter strength adjustable and adaptive rather than fixed. The filter strength varies dynamically based on coding mode, prediction mode, and sample position, allowing the filtering process to adapt to different video content characteristics and coding conditions, thereby improving versatility without excessive complexity.
3Productivity
If filtering uses only reconstructed samples within the current block, then implementation is simpler, but filtering efficiency is reduced due to limited sample availability
Solution Approach 1:
The patent uses preliminary action by generating extrapolated samples in advance before the filtering process. These samples are created by extending the current coding unit boundaries using extrapolation techniques, providing additional sample data that improves filtering efficiency without requiring complex real-time operations during the filtering stage.
Solution Approach 2:
The patent introduces an intermediary element by generating extrapolated samples that act as a bridge between the current coding unit and the filtering operation. These extrapolated samples serve as virtual extensions that provide the necessary sample data for efficient filtering without requiring actual neighboring block data, thus improving productivity while controlling complexity.
Data Source
AI summary
Systems and methods are described for video coding using adaptive Hadamard filtering of reconstructed blocks, such as coding units. In some embodiments, where Hadamard filtering might otherwise encompass samples outside the current coding unit, extrapolated samples are generated for use in the filtering. Reconstructed samples from neighboring blocks may be used in the filtering where available (e.g. in a line buffer). In some embodiments, different filter strengths are applied to different spectrum components in the transform domain. In some embodiments, filter strength is based on position of filtered samples within the block. In some embodiments, filter strength is based on the prediction mode used to code the current block.