Non-separable Secondary Transform for Image Coding Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The increasing demand for high-resolution and high-quality images, such as HD and UHD, leads to higher data transmission and storage costs due to increased information volume, necessitating a more efficient image compression technique.

Innovation Solution

A method and apparatus for enhancing image coding efficiency through a non-separable secondary transform (NSST) method, which determines and applies a suitable NSST kernel based on the intra prediction mode and block size to optimize transform coefficients, reducing data volume and improving residual coding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If high-resolution and high-quality images are transmitted or stored, then image quality is improved, but transmission cost and storage cost increase

Engineering Contradiction:
Improveimage qualityVSAvoiddata volume
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent applies parameter changes by dynamically selecting different transform kernel sizes (4x4 or 8x8) based on the block size and intra prediction mode. This adaptive parameter selection optimizes the transform process to achieve better compression efficiency, reducing the data volume required while maintaining image quality. The NSST index and kernel size are adjusted as parameters to balance between compression ratio and quality preservation.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If a non-separable secondary transform is applied to reduce data volume, then compression efficiency is improved, but transform complexity increases

Engineering Contradiction:
Improvecompression efficiencyVSAvoidtransform complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent implements dynamics by making the transform kernel size selectable based on actual processing needs. The system dynamically chooses between 4x4 and 8x8 kernels depending on the block size and prediction mode, rather than using a fixed complex transform. This dynamic adaptation simplifies the transform process while maintaining high compression efficiency, as the complexity is adjusted according to the specific coding situation.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes the transform parameter (kernel size) based on the block size and intra prediction mode. By adjusting this parameter adaptively, the system achieves high compression efficiency without consistently using the most complex transform. The parameter change allows the system to use simpler transforms when appropriate, reducing overall computational complexity while maintaining productivity.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If transform coefficients are concentrated on low-frequency components, then residual coding efficiency is improved, but coding flexibility is reduced

Engineering Contradiction:
Improveresidual coding efficiencyVSAvoidcoding flexibility
Core Design Contradiction:
ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent applies parameter changes by selecting different NSST indices and kernel sizes based on the intra prediction mode and block size. This adaptive parameter selection concentrates transform coefficients on low-frequency components to improve residual coding efficiency. The system adjusts the transform parameters dynamically to match the specific coding situation, maintaining flexibility while achieving concentration of coefficients for better compression.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3588952B1Transform method in image coding system and apparatus for same
Publication Date: 2021.04.28 LG ELECTRONICS INC
  • EP3588952B1 patent drawingFigure 1
  • EP3588952B1 patent drawingFigure 2
  • EP3588952B1 patent drawingFigure 3

AI summary

A transform method, according to the present invention, comprises the steps of: obtaining transform coefficients for a target block; determining a non-separable secondary transform (NSST) set for the target block; selecting one of a plurality of NSST kernels included in the NSST set on the basis of an NSST index; and generating modified transform coefficients by non-separable secondary-transform of the transform coefficients on the basis of the NSST kernel that has been selected, wherein the NSST set for the target block is determined on the basis of an intra-prediction mode and/or the size of the target block. According to the present invention, the amount of data transmitted, which is required for residual processing, can be reduced and residual coding efficiency can be increased.