Image Coding With Mode-Driven Subblock Transform Signaling

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image compression technologies face challenges in efficiently deriving inter prediction modes, intra prediction modes, and signaling information on geometric partitioning merge modes, leading to suboptimal encoding efficiency.

Innovation Solution

The method and apparatus allow for deriving subblock transform information based on inter prediction modes in merge or skip modes, enabling efficient signaling and improving encoding efficiency by omitting unnecessary information derivation based on intra prediction modes and geometric partitioning merge modes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If subblock transform information is derived based on inter prediction mode in merge or skip mode, then encoding precision is improved, but device complexity increases due to additional derivation steps

Engineering Contradiction:
Improveencoding precisionVSAvoiddevice complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent extracts and applies subblock transform information specifically for inter prediction modes in merge or skip mode, separating this functionality from other prediction modes. This allows the subblock transform to be applied only where beneficial, improving encoding precision for these specific modes without unnecessarily complicating the overall encoding device for all modes.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different processing approaches to different prediction modes locally. Subblock transform information is derived and applied specifically for inter prediction modes in merge or skip mode, while other modes use their traditional approaches. This localized application improves precision where needed without adding complexity everywhere.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If information on subblock transform is signaled for all prediction modes, then measurement precision is improved, but loss of information increases due to redundant signaling

Engineering Contradiction:
Improvemeasurement precisionVSAvoidloss of information
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent extracts and signals subblock transform information only for inter prediction modes in merge or skip mode, removing the redundancy of signaling this information for all prediction modes. This selective signaling maintains measurement precision for the relevant modes while reducing the overall information loss from redundant signaling.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Instead of applying subblock transform signaling universally to all prediction modes (excessive action), the patent applies it partially only to inter prediction modes in merge or skip mode where it provides the most benefit. This partial application reduces redundant information signaling while maintaining necessary precision.

Inventive Principle:
Principle #16Partial or excessive action

3Adaptability or versatility

If CIIP mode is allowed with skip mode, then adaptability is improved, but device complexity increases due to additional mode combinations

Engineering Contradiction:
ImproveadaptabilityVSAvoiddevice complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges CIIP mode with skip mode, allowing their combination to improve adaptability. By permitting CIIP mode to operate alongside skip mode, the system can adapt to different content characteristics more effectively, using intra prediction combined with inter prediction when beneficial, while maintaining the simplicity of skip mode when appropriate.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent introduces dynamic flexibility by allowing CIIP mode to be enabled or disabled based on content characteristics, even in skip mode contexts. This dynamic approach improves adaptability without permanently increasing device complexity, as the additional mode combination is available but not necessarily active in all cases.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP4645852A1Image encoding/decoding method and device, and recording medium for storing bitstream
Publication Date: 2025.11.05 LG ELECTRONICS INC
  • EP4645852A1 patent drawingFigure 1
  • EP4645852A1 patent drawingFigure 2
  • EP4645852A1 patent drawingFigure 3

AI summary

An image decoding method and device according to the present disclosure may: generate prediction samples of a current block on the basis of a predetermined inter prediction mode; derive residual samples of the current block on the basis of information on sub block transform (SBT); and reconstruct the current block on the basis of the prediction samples and residual samples of the current block. Here, at least one piece of the information on SBT may be derived on the basis of the inter prediction mode of the current block, and the information on SBT may include at least one of an SBT flag, an SBT size flag, an SBT direction flag, and an SBT location flag.