Adaptive Transform Pair Selection for Video Coding Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding methods face inefficiencies due to the use of a single preset transform pair for all transform units, leading to poor compression and decoding performance when different transform pairs yield significantly different compression effects.
Innovation Solution
The proposed method involves a coding and decoding approach that dynamically determines the transform pair for each block based on its size and type, using a transform pair index to select between DCT2, DST7, DCT8, and DST7 transform pairs, and encoding this selection using context-based adaptive binary arithmetic coding, allowing for adaptive kernel selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a single preset transform pair is used for all transform units, then device complexity is reduced, but compression efficiency deteriorates
Solution Approach 1:
The patent implements dynamic transform pair selection by introducing a transform pair index that can take different values (0-5) corresponding to different transform pairs. The decoder dynamically determines which transform pair to use based on the index value, allowing the system to adapt to different block characteristics rather than using a fixed transform pair for all blocks.
Solution Approach 2:
The patent changes the transform pair parameter based on block characteristics. By encoding transform pair index information in the bitstream and using context-based adaptive binary arithmetic coding (CABAC) with multiple context models, the system selectively changes transform parameters to match different block types and sizes, optimizing compression for each specific case.
2Loss of energy
If different transform pairs are used for different blocks, then compression efficiency is improved, but device complexity increases
Solution Approach 1:
The patent segments the transform pair selection process by creating distinct context models (context_model_0 through context_model_5) for different transform pair indices. This segmentation allows the decoder to use appropriate context models based on the transform pair being used, managing complexity through structured organization rather than monolithic processing.
Solution Approach 2:
The patent creates a universal decoding framework that handles multiple transform pairs through a single transform processing unit. The same inverse transform processing logic is used for all transform pairs, with the only variation being the specific kernel applied. This multi-functionality approach reduces complexity by reusing the same processing infrastructure for different transform types.
3Productivity
If transform pair selection is optimized for each block, then coding performance is improved, but bit-rate overhead increases
Solution Approach 1:
The patent applies partial coding by using context-based adaptive binary arithmetic coding (CABAC) that exploits statistical dependencies. Instead of encoding all transform pair index bits uniformly, the system uses context models that predict likely values based on previous blocks, coding only the necessary information to distinguish between different transform pair selections. This partial action approach reduces overhead by exploiting redundancy in transform pair selection patterns.
Data Source
AI summary
Provided is a decoding method, including: acquiring coded data of a current block; determining that a height and a width of the current block are both less than or equal to a preset threshold and the current block is a luma block; acquiring a transform pair index corresponding to the current block from the coded data, and determining the transform pair corresponding to the current block based on the transform pair index, wherein the transform pair comprises a horizontal transform kernel and a vertical transform kernel.