LFNST Index Parsing for Single Tree Image Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a need for highly efficient image/video compression techniques to effectively compress and transmit or store high-resolution and high-quality images/videos with various features, especially for immersive media like virtual reality and augmented reality content.
Innovation Solution
The method involves deriving a modified transform coefficient by determining the tree type of the current block and parsing an LFNST index based on the transform coefficient coding flag for the luma component or the application of the ISP mode, and applying an LFNST kernel to derive the modified transform coefficient.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution and high-quality image/video data is transmitted or stored using conventional methods, then image quality is maintained, but transmission cost and storage cost increase significantly
Solution Approach 1:
The patent extracts and separates transform coefficients into different categories (e.g., luma and chroma components, different frequency regions) and applies different coding strategies to each. This selective extraction allows for more efficient compression by focusing computational resources on preserving important visual information while aggressively compressing less perceptible data, thereby reducing transmission and storage costs while maintaining image quality.
Solution Approach 2:
The patent applies different coding precision and compression strategies to different regions and components of the image data. Specifically, it uses separate transform coefficient coding flags for luma and chroma components, and applies LFNST selectively based on block characteristics. This local differentiation optimizes the balance between quality preservation and compression efficiency for each region.
2Productivity
If conventional transform coding is applied uniformly to all blocks, then coding simplicity is maintained, but coding efficiency decreases for specific block types
Solution Approach 1:
The patent introduces dynamic adaptability in transform coding by using syntax elements (transform coefficient coding flags) that allow the encoder to select different coding modes based on block characteristics. The LFNST application is dynamically controlled by these flags, enabling the system to adapt its complexity to the actual content requirements of each block rather than applying a fixed uniform approach.
Solution Approach 2:
The patent changes coding parameters selectively based on block type and content characteristics. It uses separate transform coefficient coding flags for different components and applies LFNST only when beneficial. This parameter differentiation allows the system to optimize coding efficiency for specific block types without uniformly increasing complexity across all blocks.
3Measurement precision
If LFNST is applied to all blocks regardless of type, then transformation accuracy is improved, but processing overhead increases
Solution Approach 1:
The patent performs preliminary assessment of block characteristics before applying LFNST. The transform coefficient coding flags are evaluated first to determine whether LFNST should be applied. This preliminary action prevents unnecessary LFNST processing on blocks where it would not provide benefit, thereby reducing processing overhead while maintaining transformation accuracy where needed.
Solution Approach 2:
The patent applies LFNST selectively to only those blocks where it provides meaningful improvement, rather than applying it universally. By using the transform coefficient coding flags to identify suitable candidates, the system performs partial action on a subset of blocks, avoiding the excessive processing overhead that would result from applying LFNST to all blocks without discrimination.
Data Source
AI summary
An image decoding method according to the present document may comprise a step for deriving modified transform coefficients, wherein the step for deriving modified transform coefficients may comprise the steps of: determining whether the tree type of the current block is a single tree; parsing an LFNST index when the tree type is the single type, the LFNST index being parsed on the basis of a transform coefficient coding flag value for a luma component of the current block, or on the basis of whether an ISP mode is applicable to the luma component; and deriving the modified transform coefficients by applying an LFNST kernel derived on the basis of the LFNST index.