Video Secondary Transform Coding with Intra-Mode Kernel Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution and high-quality images/videos, particularly in immersive media formats like VR and AR, has led to a need for more efficient image/video compression techniques to reduce transmission and storage costs.
Innovation Solution
An image coding method and apparatus based on a reduced secondary transform (RST) that optimizes the transformation kernel matrix and changes the array of transform coefficients according to the intra prediction mode, enhancing coding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional transform methods are used for high-resolution images/videos, then image quality is maintained, but transmission cost and storage cost increase due to increased bit amount
Solution Approach 1:
The patent applies parameter changes by modifying the transform kernel matrix based on intra prediction modes. Different kernel matrices are selected according to the prediction mode (e.g., planar, DC, or angular modes), which optimizes the transform coefficients for different image characteristics. This adaptive approach improves compression efficiency while maintaining image quality by better matching the transform to the local image structure.
Solution Approach 2:
The patent introduces dynamics by making the secondary transform adaptive to intra prediction modes. The transform kernel matrix is dynamically selected and adjusted based on the prediction mode used for the current block, allowing the transformation to adapt to different image regions and prediction directions, thereby improving compression performance without sacrificing quality.
2Productivity
If secondary transform is applied to all blocks, then coding efficiency improves, but computational complexity and memory requirements increase
Solution Approach 1:
The patent applies local quality by differentiating the secondary transform application based on intra prediction modes. Instead of uniformly applying the transform to all blocks, the patent selectively applies different kernel matrices or skips the secondary transform for certain prediction modes (e.g., planar or DC modes where it may be redundant). This localized approach improves coding efficiency where beneficial while reducing computational complexity where unnecessary.
Solution Approach 2:
The patent implements partial action by conditionally applying the secondary transform only to specific blocks based on their intra prediction mode. Rather than applying the transform universally, the patent selectively applies it to blocks where it provides the most benefit, thereby improving overall coding efficiency while limiting the increase in computational complexity to only the necessary portions.
3Productivity
If transform kernel matrix is optimized for specific intra prediction modes, then transform efficiency improves, but device complexity increases due to multiple kernel matrices
Solution Approach 1:
The patent applies universality by designing a set of transform kernel matrices that can serve multiple intra prediction modes. The same kernel matrix may be used for different prediction modes that share similar characteristics, reducing the total number of unique kernels needed while still providing optimized transformation for various image structures. This multi-functional approach improves transform efficiency without proportionally increasing device complexity.
Data Source
AI summary
A video transform method includes: receiving a quantized transform coefficient for a target block and a transform index for a non-separated secondary transform; inverse-quantizing the quantized transform coefficient; deriving an input transform coefficient size indicating the length of the inverse-quantized transform coefficient to which the non-separated secondary transform has been applied, an output transform coefficient size indicating the length of a modified transform coefficient to which the non-separated secondary transform has been applied, and a transform set mapped to the intra mode of the target block, when the transform index does not indicate that the non-separated secondary transform is not performed; and deriving the modified transform coefficient on the basis of a matrix operation on a transform kernel matrix in the transform set indicated by the transform index, and a transform coefficient list corresponding to the input transform coefficient size.