CCLM Chroma Prediction Using Right-Shift Parameter Derivation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing moving image encoding and decoding technologies, such as CCLM prediction, require large memory and complex operations due to the use of integer operations and table lookups, leading to increased hardware scale and complexity.
Innovation Solution
A simplified CCLM prediction method that uses right-shift operations to derive CCLM prediction parameters, reducing the complexity of multiplication and memory requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If integer operations and table lookups are used to derive linear prediction parameters in CCLM prediction, then prediction accuracy is maintained, but memory requirements and hardware complexity increase significantly
Solution Approach 1:
The patent extracts and removes the complex table lookup mechanism from the CCLM prediction parameter derivation process. Instead of using pre-computed tables that require large memory, the invention directly computes the linear prediction parameters using simplified arithmetic operations, thereby eliminating the need for large lookup tables while maintaining prediction accuracy.
Solution Approach 2:
The patent replaces the mechanical table lookup operation with arithmetic computations. The linear prediction parameters are derived through direct mathematical calculations involving pixel value differences and gradient computations, substituting the mechanical memory access and table indexing operations with computational logic that reduces hardware complexity.
2Measurement precision
If large bit-width values are used in product calculations for CCLM prediction parameters, then prediction precision is maintained, but hardware scale and complexity increase
Solution Approach 1:
The patent changes the parameters used in the product calculations by introducing scaled-down intermediate values. Instead of directly multiplying large bit-width pixel values, the invention uses divided or shifted versions of these values that maintain the essential prediction information while reducing the bit-width requirements, thereby decreasing hardware scale.
Solution Approach 2:
The patent applies partial precision in intermediate calculation steps. By using approximate values with reduced bit-width for intermediate products and only maintaining full precision for the final prediction parameters, the invention achieves adequate prediction precision without requiring full precision throughout the entire calculation chain, thus reducing hardware scale.
3Reliability
If comprehensive table lookups are performed for deriving CCLM prediction parameters, then prediction quality is improved, but processing time and complexity increase
Solution Approach 1:
The patent extracts the time-consuming table lookup operation from the prediction parameter derivation process. By removing the dependency on pre-computed tables and using direct arithmetic calculations, the invention eliminates the memory access delays associated with table lookups, thereby reducing processing time while maintaining prediction quality through accurate computational methods.
4Measurement precision
If large memory is allocated for table storage in CCLM prediction, then parameter derivation accuracy is maintained, but device complexity and resource usage increase
Solution Approach 1:
The patent extracts and eliminates the large memory allocation for table storage by replacing the table-based approach with direct computational derivation. The linear prediction parameters are calculated on-the-fly using arithmetic operations on pixel values, removing the need for storing large lookup tables in memory, thereby maintaining parameter accuracy while dramatically reducing memory usage.
Data Source
AI summary
A moving image decoding method for deriving a prediction image of a chrominance image by using a luminance image is provided. The method derives a first luminance value and a first chrominance value corresponding to a first position, and derives a second luminance value and a second chrominance value corresponding to a second position. The method further derives a first difference value between the first and second luminance values, and derives a second difference value between the first and second chrominance values. The method sets a shift value to a first threshold based on a comparison between the first threshold and a third value, being calculated by adding a first specified value to a first value, derived by using the first difference value, and then subtracting a second value, derived by using the second difference value, from an addition result, and derives the prediction image by using the shift value.