Video Encoder Secondary-Transform Quantization for Image Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding standards face challenges in efficiently encoding images while maintaining subjective image quality, particularly when secondary transforms are applied.
Innovation Solution
An encoder that performs a primary transform on image residuals, determines whether a secondary transform is needed, and applies different quantizations based on the application of the secondary transform, using distinct quantization matrices for primary and secondary spaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a secondary transform is applied to the current block, then coding efficiency is improved, but subjective image quality decreases
Solution Approach 1:
The patent applies different quantization matrices to different transform spaces (primary vs. secondary). When a secondary transform is applied, a second quantization matrix is used that is specifically tailored for the secondary transform space, rather than using a universal quantization matrix. This localized optimization allows the system to achieve better coding efficiency in the secondary transform path while maintaining image quality through space-appropriate quantization characteristics.
Solution Approach 2:
The patent changes the quantization parameters based on the transform type being applied. By determining whether a secondary transform is applied to the current block, the system dynamically selects between a first quantization matrix (for primary transform only) and a second quantization matrix (for secondary transform applied). This parameter adaptation allows optimization of coding efficiency for secondary transforms while controlling quality degradation through appropriate quantization strength selection.
2Productivity
If different quantization matrices are used for primary and secondary spaces, then coding efficiency is enhanced, but device complexity increases
Solution Approach 1:
The patent implements a dynamic quantization system where the selection of quantization matrix depends on the transform path taken. The system determines whether a secondary transform is applied and dynamically switches between using a first quantization matrix (primary transform only) or a second quantization matrix (secondary transform applied). This dynamic adaptation enables coding efficiency enhancement through specialized matrices while managing complexity through conditional logic rather than always processing with multiple matrices.
Solution Approach 2:
The patent applies different quantization matrices to different transform spaces (primary vs. secondary). When a secondary transform is applied, a second quantization matrix is used that is specifically tailored for the secondary transform space, rather than using a universal quantization matrix. This localized optimization allows the system to achieve better coding efficiency in the secondary transform path while maintaining image quality through space-appropriate quantization characteristics.
Data Source
AI summary
An encoder which encodes a current block to be encoded in an image includes: a transformer which performs a primary transform from residuals of the current block to primary coefficients, determines whether to apply a secondary transform to the current block, and performs the secondary transform from the primary coefficients to secondary coefficients when the secondary transform is applied; a quantizer which calculates quantized primary coefficients by performing a first quantization on the primary coefficients when a secondary transform is not applied, and calculates quantized secondary coefficients by performing a second quantization different from the first quantization on the secondary coefficients when the secondary transform is applied; and an entropy encoder which generates an encoded bitstream by encoding either quantized primary coefficients or quantized secondary coefficients.