CC-ALF Filter Coefficient Extension for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Cross-Component Adaptive Loop Filter (CC-ALF) designs in video coding face challenges in maintaining filtering precision while reducing hardware complexity, leading to increased bit-depth rate (BD-rate) and potential quality degradation due to restricted coefficient values.
Innovation Solution
The proposed solution extends the allowed CC-ALF coefficient values to include a broader range of values beyond pure power-of-two multiples, using shifts and additions to calculate residual corrections, allowing for improved filtering without reintroducing multiplication operations, thus maintaining low complexity and enhancing filtering precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If CC-ALF coefficient values are restricted to pure power-of-two multiples, then hardware complexity is reduced, but filtering precision deteriorates leading to increased BD-rate
Solution Approach 1:
The patent applies local quality by differentiating between two types of coefficient values: those that are pure power-of-two multiples (used where hardware simplicity is critical) and those that are sums of power-of-two multiples (used where improved precision is needed). This allows the system to locally optimize for precision in specific filter positions while maintaining overall low complexity through the structured form of allowed values.
Solution Approach 2:
The patent introduces dynamics by allowing the coefficient values to be selected from an extended set that includes both pure power-of-two multiples and sums thereof. This dynamic expansion of the coefficient value space enables the system to adapt between complexity and precision requirements based on the specific filtering needs at different positions, rather than being statically constrained to a single value type.
2Measurement precision
If CC-ALF coefficient values are extended beyond pure power-of-two multiples, then filtering precision is improved, but hardware complexity increases
Solution Approach 1:
The patent changes the parameter space of allowed coefficient values from exclusively pure power-of-two multiples to include sums of power-of-two multiples. This parameter expansion enables higher filtering precision by providing more granular control over filter characteristics, while the structured nature of the extended values (still based on power-of-two building blocks) helps manage the complexity increase.
Solution Approach 2:
The patent segments the coefficient value representation into discrete power-of-two components that can be summed. This segmentation allows the hardware to implement complex coefficient values through a series of simpler operations (adding shifted values) rather than requiring full multiplication capability, thus managing complexity while achieving extended precision.
3Adaptability or versatility
If multiplication operations are reintroduced in CC-ALF, then filtering flexibility is improved, but hardware complexity increases
Solution Approach 1:
The patent substitutes multiplication operations with addition and bit-shifting operations. Instead of using full multiplication hardware to achieve filtering flexibility, the system represents coefficient values as sums of power-of-two multiples, which can be implemented through addition of shifted versions of the input signal. This substitution maintains adaptability while avoiding the complexity of multiplication hardware.
Solution Approach 2:
The patent applies partial action by providing filtering flexibility for common cases through the extended coefficient set (sums of power-of-two multiples) without implementing full multiplication capability. The system achieves sufficient flexibility for most filtering scenarios using the extended additive approach, reserving full multiplication-only implementations for exceptional cases where maximum flexibility is absolutely required.
Data Source
AI summary
There is provided a method for encoding or decoding an image. The method comprises obtaining a first luma sample value, L1, associated with the image. The method comprises obtaining a second luma sample value, L2, associated with the image. The method further comprises obtaining a first luma delta value, ΔL1, wherein ΔL1=L2−L1. The method comprises obtaining a first product, P1, using ΔL1 and a first coefficient value, C1, wherein P1=(C1)(ΔL1). The method comprises calculating a first residual correction value, ΔI1 using P1 and a set of other products. The method comprises filtering an unfiltered chroma value, RC, associated with the image using the first residual correction value, ΔI1, thereby producing a filtered chroma value RFC associated with the image.