LFNST Matrix Application in Image Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing demand for high-resolution and high-quality images/videos, such as 4K and 8K ultra high definition, leads to higher bit rates, resulting in increased transmission and storage costs. Additionally, the need for efficient compression techniques is exacerbated by the growing interest in immersive media like virtual and augmented reality.
Innovation Solution
The proposed solution involves an image coding method and apparatus that utilizes Low-Frequency Non-Separable Transform (LFNST) to enhance coding efficiency. This method includes deriving transform coefficients, determining the existence of significant coefficients in specific regions, parsing an LFNST index, applying an LFNST matrix to modify transform coefficients, and performing inverse primary transforms to generate reconstructed pictures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional image coding methods are used for high-resolution images, then image quality is maintained, but transmission and storage costs increase due to higher bit rates
Solution Approach 1:
The patent applies Low-Frequency Non-Separable Transform (LFNST) with different transform matrices (4×4, 8×8, 16×16) based on block size and content characteristics. By changing the transform parameters adaptively, the coding efficiency is improved while maintaining image quality, thereby reducing the bit rate required for transmission and storage.
Solution Approach 2:
The patent determines whether to apply LFNST and which transform matrix to use based on local characteristics of the image block, such as the presence of significant coefficients in specific regions. This localized approach optimizes compression for each block independently, improving overall compression efficiency while preserving important image details.
2Productivity
If LFNST is applied to all blocks, then compression efficiency improves, but computational complexity and device requirements increase
Solution Approach 1:
The patent does not apply LFNST to all blocks unconditionally. Instead, it selectively applies LFNST based on criteria such as the presence of significant coefficients in specific regions and block size. This partial application approach achieves compression efficiency improvements while avoiding the computational overhead of processing every block with LFNST.
Solution Approach 2:
The patent divides the transform process into segments: first applying a primary transform, then conditionally applying LFNST only to certain blocks that meet specific criteria. This segmentation allows the system to achieve compression benefits where needed while reducing overall computational complexity by skipping LFNST for blocks that don't benefit from it.
3Speed
If transform index coding is simplified, then encoding speed increases, but transform precision and image quality may deteriorate
Solution Approach 1:
The patent uses dynamic decision-making for transform index selection based on block characteristics. The encoder determines whether to apply LFNST and which matrix to use based on real-time analysis of significant coefficient positions and block properties. This dynamic approach allows efficient encoding by avoiding unnecessary transforms while maintaining precision where needed.
Solution Approach 2:
The patent performs preliminary analysis of the transform coefficients to determine significant coefficient positions before selecting the transform matrix. This preliminary action allows the system to make informed decisions about which LFNST matrix to apply, ensuring transform precision is maintained while avoiding the need to evaluate all possible matrices, thus improving encoding speed.
Data Source
AI summary
An image decoding method, according to the present document, may comprise the steps of: deriving transform coefficients for a current block on the basis of residual information; determining whether a significant coefficient is present in a second region excluding a first region in the top-left end of the current block; parsing a LFNST index from a bitstream if the significant coefficient is not present in the second region; deriving modified transform coefficients by applying a LFNST matrix, derived on the basis of the LFNST index, to transform coefficients of the first region; and deriving residual samples of the current block on the basis of an inverse primary transform of the modified transform coefficients.