LFNST Matrix Design for Low-Complexity Image Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution, high-quality images and videos, particularly in virtual reality and augmented reality, requires a highly efficient image/video compression technique to minimize transmission and storage costs while considering computational complexity.
Innovation Solution
An image coding method and apparatus utilizing a low-frequency non-separable transform (LFNST) matrix, which derives modified transform coefficients and generates residual samples based on the LFNST index, optimizing coding performance and minimizing complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution, high-quality image/video compression is applied, then image quality and resolution are improved, but transmission and storage costs increase
Solution Approach 1:
The patent applies Low-Frequency Non-Separable Transform (LFNST) with specifically designed kernel matrices to transform transform coefficients, changing the mathematical parameters of the compression process. This enables more efficient representation of image data, achieving high quality compression with reduced bitrates, thereby lowering transmission and storage costs while maintaining image quality
2Productivity
If LFNST is applied to improve coding performance, then compression efficiency is improved, but computational complexity increases
Solution Approach 1:
The patent applies LFNST selectively based on block size conditions (e.g., 16×16, 32×32, 64×64 blocks) rather than universally to all blocks. The transform is applied only when it provides beneficial compression performance, avoiding unnecessary computational overhead for blocks where LFNST would not improve coding efficiency, thus balancing complexity and performance
3Adaptability or versatility
If LFNST matrix derivation is performed for all block sizes, then coding flexibility is improved, but processing complexity increases
Solution Approach 1:
The patent derives LFNST kernel matrices with specific dimensions (e.g., 96×32, 32×96) tailored to specific block size conditions. Different kernel matrix configurations are prepared for different block sizes (16×16, 32×32, 64×64), allowing the system to adapt to local requirements of each block type while avoiding the complexity of preparing all possible matrix configurations for all block sizes
Data Source
AI summary
An image decoding method according to this document comprises the steps of: deriving an LFNST matrix for a current block on the basis of an LFNST index derived from LFNST index information and an LFNST set index; deriving modified transform coefficients on the basis of transform coefficients and the LFNST matrix; and generating residual samples for the current block on the basis of the modified transform coefficients, wherein when the width or height of the current block has a value of 16 and both the width and height have a value of 16 or more, the LFNST matrix may be derived as a 96×32 dimensional matrix. Therefore, coding performance achievable by the LFNST can be maximized within the implementation complexity permitted in forthcoming standards.