Model-Based ALF Taps for Accurate Chroma Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face challenges in efficiently improving video quality through adaptive loop filtering, particularly in handling cross-component filtering for chroma components, which can amplify artifacts and require complex filtering processes.
Innovation Solution
Introduce model-based taps for adaptive loop filtering (ALF) using interpolation and cross-component models to generate tap inputs, deriving samples at non-integer positions and applying adaptive filters based on historical data and neighboring reference areas.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If cross-component filtering is applied to chroma components, then video quality is improved, but artifacts are amplified and filtering complexity increases
Solution Approach 1:
The patent divides the filtering process into separate components: luma component filtering and chroma component filtering are handled independently through the cross-component model. The chroma filtering is further segmented by deriving model-based taps from luma samples, allowing the complex chroma filtering to be broken down into simpler operations that reference the already-processed luma component.
Solution Approach 2:
The patent introduces model-based taps as an intermediary mechanism that connects the luma component to the chroma component filtering process. These taps are derived from luma samples using a cross-component model and serve as intermediate values that simplify the chroma filtering operation, reducing direct complexity while maintaining filtering accuracy.
2Measurement precision
If adaptive loop filtering is applied to improve video quality, then filtering accuracy is improved, but processing complexity and computational load increase
Solution Approach 1:
The patent applies preliminary action by first filtering the luma component through the adaptive loop filter before using its samples to derive model-based taps for chroma filtering. This preliminary processing of the luma component creates ready-to-use reference data that simplifies subsequent chroma filtering operations, reducing overall processing complexity while maintaining accuracy.
Solution Approach 2:
The patent changes the parameter representation by deriving model-based taps from luma samples rather than directly processing chroma samples. This parameter transformation allows the chroma filtering to leverage the already-optimized luma filtering results, reducing computational load while preserving filtering accuracy through the cross-component relationship.
3Measurement precision
If model-based taps are introduced for chroma filtering, then filtering accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent uses copying by deriving model-based taps that replicate the essential characteristics of chroma samples from luma samples through the cross-component model. Instead of directly processing chroma samples, the system creates copies of relevant luma sample information that capture the chroma component's filtering needs, simplifying the computational process while maintaining accuracy.
Data Source
AI summary
Method and apparatus to generate cross-component model-based taps for ALF. According to this method, input data associated with a current block including a first-colour block and a second-colour block are received, where the first-colour block includes first-colour samples and the second-colour block includes second-colour samples. One or more target second-colour samples are derived according to a Cross-Component Model (CCM) applied to one or more CCM-input first-colour samples or deriving said one or more target second-colour samples at one or more non-integer positions by applying one or more interpolation filters to one or more interpolation-input first-colour or second-colour samples. One or more filtered second-colour samples are generated by applying target ALF (Adaptive Loop Filter) using filter input samples comprising one or more filter-input second-colour samples and said one or more target second-colour samples.