LFNST Quantization Layout for Video Compression Efficiency
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, necessitates a more efficient compression technique to reduce transmission and storage costs.
Innovation Solution
The implementation of a low-frequency non-separable transform (LFNST) in video/image coding systems, which includes quantization and scaling based on the position of coefficients, determines whether to apply scaling list data, and enhances compression efficiency through quantization parameter configuration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If conventional video/image coding is used for high-resolution content, then transmission and storage costs increase, but image quality is maintained
Solution Approach 1:
The patent applies Low-Frequency Non-Separable Transform (LFNST) as a parameter change in the transform domain. By transforming low-frequency coefficients using a non-separable kernel, the patent achieves better energy compaction and compression efficiency while maintaining perceptual image quality, thus reducing transmission and storage costs without sacrificing image quality
2Productivity
If LFNST is applied to all coefficients, then compression efficiency improves, but computational complexity increases
Solution Approach 1:
The patent applies LFNST selectively rather than uniformly across all coefficients. By applying the non-separable transform only to specific low-frequency coefficient regions (typically the top-left 4x4 or 8x8 block), the patent achieves improved compression efficiency for the most important frequency components while limiting the increase in computational complexity to only those localized regions
3Productivity
If quantization parameter is configured based on coefficient position, then compression efficiency improves, but processing complexity increases
Solution Approach 1:
The patent divides the transform coefficient block into multiple regions based on frequency position, with different quantization parameter configurations applied to different segments. This segmentation allows the patent to apply stronger quantization to less important high-frequency regions while preserving more detail in low-frequency regions, improving overall compression efficiency with manageable processing complexity
Data Source
AI summary
According to the disclosure of the present document, a quantization parameter for a quantized transform coefficient present at a DC position within a current block can have the smallest value among quantization parameters which are applied to quantized transform coefficients within the current block, on the basis that an LFNST is applied to the current block. Therefore, coding performance capable of being brought by the LFNST can be maximized within the implementation complexity permitted in forthcoming standards.