Training-Based Video Coding with Mode-Adaptive Transform Kernels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in achieving high compression efficiency and accuracy, particularly in selecting appropriate transform kernels and intra prediction modes, which affect coding quality and computational complexity.
Innovation Solution
Adaptive selection of transform kernels based on decoder-side intra mode derivation (DIMD) and intra mode replacement sets, combined with non-separable primary transforms, to improve coding accuracy and reduce computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional transform kernels are used for all blocks, then device complexity is reduced, but coding accuracy deteriorates
Solution Approach 1:
The patent applies dynamics by making the transform kernel selection adaptive rather than fixed. Different transform kernels are dynamically selected based on the prediction mode (intra or inter) and the specific block characteristics. This allows the system to optimize coding accuracy for each block while managing complexity through rule-based selection criteria.
Solution Approach 2:
The patent implements local quality by applying different transform kernels to different blocks based on their specific characteristics. Intra blocks use one set of kernels while inter blocks use another set, and further differentiation is made based on prediction mode and block size. This localized optimization improves overall coding accuracy without requiring every possible block to use the most complex kernel.
2Manufacturing precision
If multiple transform kernel sets are used, then coding accuracy is improved, but device complexity increases
Solution Approach 1:
The patent segments the transform kernel selection into distinct categories based on prediction mode (intra vs. inter) and further subdivides them by block size and characteristics. This segmentation organizes the multiple kernel sets into manageable groups with clear selection rules, reducing the cognitive load and computational overhead of kernel selection while maintaining high coding accuracy.
3Manufacturing precision
If adaptive transform kernel selection is implemented, then coding accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-defining the transform kernel selection rules and categories before encoding/decoding occurs. The rules for selecting kernels based on prediction mode, block size, and characteristics are established in advance, allowing the encoder/decoder to quickly determine the appropriate kernel without complex real-time calculations, thus maintaining high speed despite adaptive selection.
4Manufacturing precision
If decoder-side intra mode derivation is used, then coding accuracy is improved, but device complexity increases
Solution Approach 1:
The patent implements self-service by having the decoder independently derive intra modes using the same reference samples and rules as the encoder. The decoder uses its own reconstructed blocks and reference picture data to perform the derivation without requiring additional signaling or complex external information, simplifying the implementation while maintaining accuracy.
Data Source
AI summary
An example method of video coding includes receiving a video bitstream comprising a current block and identifying a first prediction mode for the current block. The method also includes, when the first prediction mode is a particular prediction mode, selecting a first set of transform kernels as transform kernels for the current block, and, when the first prediction mode is not the particular prediction mode, selecting a second set of transform kernels as the transform kernels for the current block. The method further includes applying a transform for the current block using the transform kernels.