Image Decoding Candidate List Generation for Inter Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image compression technologies face challenges in efficiently generating candidate lists for inter prediction, particularly in deriving motion information for temporal candidates and adding various candidates to these lists.

Innovation Solution

An image decoding method and device that generate a candidate list for a current block, derive motion information based on spatial, temporal, or derived candidates, and perform inter prediction using this motion information. The method involves determining top-right and bottom-left blocks adaptively based on block sizes and shapes, and selectively adding derived candidates to the list.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional inter prediction methods are used to generate candidate lists, then the encoding/decoding process is simpler, but the prediction accuracy is insufficient

Engineering Contradiction:
Improveprediction accuracyVSAvoidcandidate list generation complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The candidate list is segmented into multiple types (spatial candidates, temporal candidates, and derived candidates), with each type obtained through different methods. This segmentation allows the system to explore diverse motion information sources without overwhelming complexity, as each segment has a specific derivation approach

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Derived candidates act as intermediaries between spatial and temporal candidates. They are obtained by adaptively combining motion information from spatial neighboring blocks and temporal neighboring blocks, serving as a bridge that leverages the strengths of both approaches while mitigating their individual limitations

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If motion information from multiple sources is combined to derive candidates, then the reliability of inter prediction is improved, but the computational complexity increases

Engineering Contradiction:
Improveinter prediction reliabilityVSAvoidmotion information processing complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The method dynamically adapts the derivation process based on the availability of motion information from spatial and temporal neighboring blocks. The derivation adaptively adjusts which candidates to generate and how to combine motion information, allowing the system to optimize reliability while managing complexity based on actual scene conditions

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The method changes parameters such as the number of candidates in the list and the derivation approach based on the availability and quality of motion information from different sources. By adjusting these parameters adaptively, the system improves prediction reliability without incurring excessive computational complexity in all cases

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If a comprehensive candidate list with multiple candidate types is generated, then the prediction quality is enhanced, but the encoding overhead increases

Engineering Contradiction:
Improveprediction qualityVSAvoidencoding overhead
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The derived candidate derivation method serves multiple functions: it enhances prediction quality by providing additional motion information sources, maintains compatibility with existing spatial and temporal candidate methods, and adaptively adjusts to different scene types. This multi-functionality allows comprehensive prediction quality enhancement without proportionally increasing encoding overhead

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4554219A1Image encoding/decoding method and device, and recording medium having bitstream stored therein
Publication Date: 2025.05.14 LG ELECTRONICS INC
  • EP4554219A1 patent drawingFigure 1
  • EP4554219A1 patent drawingFigure 2
  • EP4554219A1 patent drawingFigure 3

AI summary

An image encoding/decoding method and device according to the present disclosure may generate a candidate list of a current block and perform inter prediction of the current block on the basis of motion information of any one of multiple candidates belonging to the candidate list. The multiple candidates may include at least one of spatial candidates, temporal candidates, or derived candidates.