Bi-Prediction Weight Index Derivation for Affine Merge Coding

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The increasing demand for high resolution and high quality images/videos, such as 4K and 8K Ultra High Definition, along with the growth of virtual reality and augmented reality content, necessitates a highly efficient image/video compression technique to reduce transmission and storage costs while effectively handling diverse image/video characteristics.

Innovation Solution

The method involves deriving weight index information for generating prediction samples using affine merge candidates, including control point motion vectors, to enhance image coding efficiency during inter prediction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If high resolution and high quality image/video data are transmitted, then image quality and detail are improved, but transmission cost and storage cost increase

Engineering Contradiction:
Improveimage qualityVSAvoidtransmission cost
Core Design Contradiction:
Manufacturing precisionVSLoss of energy

Solution Approach 1:

The patent applies parameter changes by using affine motion models to transform motion vectors between different reference frames and pictures. This allows the system to adapt motion information to different temporal and spatial parameters, enabling efficient prediction of high-resolution images from lower-resolution references, thus improving image quality while reducing transmission and storage costs

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent implements dynamics by dynamically selecting between different motion models (affine motion model vs. conventional motion model) based on the inter prediction mode. The system adapts its behavior based on the specific coding situation, using affine motion models when beneficial for bi-directional prediction and conventional models otherwise, optimizing the balance between compression efficiency and computational complexity

Inventive Principle:
Principle #15Dynamics

2Ease of operation

If conventional motion models are used for inter prediction, then processing is simpler, but prediction accuracy for bi-directional prediction is insufficient

Engineering Contradiction:
Improveprocessing simplicityVSAvoidprediction accuracy
Core Design Contradiction:
Ease of operationVSMeasurement precision

Solution Approach 1:

The patent implements dynamics by dynamically selecting between different motion models (affine motion model vs. conventional motion model) based on the inter prediction mode. The system adapts its behavior based on the specific coding situation, using affine motion models when beneficial for bi-directional prediction and conventional models otherwise, optimizing the balance between compression efficiency and computational complexity

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent applies parameter changes by using affine motion models to transform motion vectors between different reference frames and pictures. This allows the system to adapt motion information to different temporal and spatial parameters, enabling efficient prediction of high-resolution images from lower-resolution references, thus improving image quality while reducing transmission and storage costs

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If weight index information is derived for each candidate in merge candidate list, then prediction accuracy is improved, but processing complexity increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidprocessing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent applies local quality by deriving weight index information selectively based on the specific candidate type in the merge candidate list. Different weight index derivation methods are applied to different candidates (e.g., using motion vector differences for affine merge candidates vs. using reference picture indices for temporal merge candidates), optimizing prediction accuracy for each local case while avoiding unnecessary complexity for all candidates

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent implements preliminary action by pre-defining the weight index derivation methods for different candidate types before the actual prediction process. The system has prepared multiple derivation approaches (using motion vector differences, using reference picture indices, using a combination) and selects the appropriate pre-defined method based on the candidate type, avoiding the need for complex real-time decision-making

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP3985980B1Image decoding method and image encoding method for deriving weight index information for generation of prediction sample
Publication Date: 2026.01.21 LG ELECTRONICS INC
  • EP3985980B1 patent drawingFigure 1
  • EP3985980B1 patent drawingFigure 2
  • EP3985980B1 patent drawingFigure 3

AI summary

According to the disclosure of the present document, when the inter prediction type of a current block indicates bi-prediction, weight index information for candidates in a merge candidate list or a sub-block merge candidate list can be derived, and thus coding efficiency can be increased.