Virtual Reference Pictures for Residual Reduction in Inter-Picture Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video coding technologies face challenges in reducing residual data after inter-prediction, leading to increased coded bit size despite improvements in coding efficiency through motion compensation.

Innovation Solution

Generating more accurate reference pictures using neural-network processing to improve video compression by creating virtual reference pictures from decoded reference pictures, which are then used for inter-picture prediction.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If motion compensation is used for inter-picture prediction, then coding efficiency is improved, but residual data increases leading to larger coded bit size

Engineering Contradiction:
Improvecoding efficiencyVSAvoidresidual data
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The patent generates virtual reference pictures in advance from existing decoded reference pictures before they are needed for prediction. This preliminary generation of enhanced reference pictures allows the system to start with better prediction material, reducing the residual that needs to be encoded later in the process.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces virtual reference pictures as an intermediary element between the decoded reference pictures and the current picture being encoded. These virtual reference pictures act as a mediator that bridges the gap, providing enhanced prediction material that more closely resembles the actual current picture content, thereby reducing residual data.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If more accurate reference pictures are generated using neural-network processing, then prediction accuracy is improved, but device complexity increases

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

Solution Approach 1:

The patent creates virtual reference pictures by copying and transforming existing decoded reference pictures through neural network processing. Instead of requiring completely new complex processing systems, the approach replicates and enhances existing reference materials, achieving improved accuracy while managing complexity through reuse of existing picture data.

Inventive Principle:
Principle #26Copying

3Loss of information

If virtual reference pictures are generated from decoded reference pictures, then residual data is reduced, but processing time increases

Engineering Contradiction:
Improveresidual dataVSAvoidprocessing time
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The virtual reference pictures are generated in advance during the decoding of reference pictures, before they are needed for the current picture encoding. This preliminary action allows the system to prepare enhanced prediction material without adding time pressure to the critical encoding path, reducing residual data efficiently.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4038579B1Method and apparatus for inter-picture prediction with virtual reference picture for video coding
Publication Date: 2025.10.01 TENCENT AMERICA LLC
  • EP4038579B1 patent drawingFigure 1
  • EP4038579B1 patent drawingFigure 2
  • EP4038579B1 patent drawingFigure 3

AI summary

A method of performing inter-picture prediction of a current picture of a video sequence is performed by at least one processor and includes determining whether the current picture uses a virtual reference picture, and based on the current picture being determined to use the virtual reference picture, generating, for the current picture, the virtual reference picture from neighboring reference pictures that are nearest neighbors to the virtual reference picture, storing the generated virtual reference picture in a decoded picture buffer, and adding the generated virtual reference picture to an active reference picture list. The method further includes performing the inter-picture prediction of the current picture, based on the active reference picture list to which the virtual reference picture is added.