Chroma Intra Prediction Direction Derivation for Lower VVC Decoding Complexity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing chroma intra prediction modes in Versatile Video Coding (VVC) suffer from high complexity, sequence dependence, and prediction inaccuracy due to the use of fixed modes like HOR, VER, and 66, which are not well adapted to all sequences, and the Multiple Direct Mode Signaling (MDMS) requires extensive comparisons and high dependence on neighboring blocks.
Innovation Solution
A method to determine a prediction direction by acquiring a Direct Mode (DM) and deriving an index number based on an offset N and index number M, adjusting the DM derived mode to improve prediction accuracy by modifying the index numbers according to specific conditions and priorities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Multiple Direct Mode Signaling (MDMS) is used to construct candidate list, then prediction accuracy is improved, but decoding complexity increases due to large comparison count
Solution Approach 1:
The patent extracts and removes fixed modes (HOR, VER, 66) from the candidate list construction process. By taking out these predetermined modes that cause high sequence dependence and unnecessary comparisons, the method reduces decoding complexity while maintaining prediction accuracy through dynamic DM-based mode selection only when beneficial.
Solution Approach 2:
The patent introduces dynamic adaptation by using Direct Mode (DM) with derived modes instead of static fixed modes. The candidate list construction becomes dynamic based on actual block characteristics and neighboring block information, allowing the system to adapt to different sequences and blocks rather than applying fixed patterns universally.
2Device complexity
If fixed modes (HOR, VER, 66) are used for default candidate angular mode, then device complexity is reduced, but prediction accuracy deteriorates due to high sequence dependence
Solution Approach 1:
The patent changes the parameters used for mode selection from fixed predetermined values to dynamic values derived from Direct Mode (DM). By deriving modes based on actual block characteristics and adjusting the candidate list dynamically, the system achieves better prediction accuracy without significantly increasing complexity, as the derivation process is more efficient than full MDMS comparison.
3Measurement precision
If DM is used for chroma block prediction, then prediction accuracy is improved for local texture features, but reliability decreases when multiple collocated luma blocks exist due to inability to reflect global features
Solution Approach 1:
The patent makes the prediction system universal by integrating multiple information sources: local texture features from DM, global features from multiple collocated luma blocks, and neighboring block information. The method can adaptively select and combine different prediction approaches depending on the block characteristics, making it reliable for both local and global feature scenarios.
Solution Approach 2:
The patent introduces an intermediary selection mechanism that chooses between different prediction approaches. Based on block characteristics and the availability of collocated luma blocks, the system mediates between using DM for local texture prediction and using broader contextual information for global feature prediction, ensuring reliability across different scenarios.
Data Source
AI summary
Provided are a method for determining a prediction direction, a decoder and a computer storage medium. The method is implemented by a decoder. The method includes that: a Direct Mode (DM) in a chroma intra prediction mode of a block to be decoded is acquired (S101); and an index number of a prediction direction in a DM derived mode is determined based on an offset N and an index number M of a prediction direction in the DM to determine the DM derived mode (S102).