Bidirectional Inter Prediction Using Shared Motion Vector Difference

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video encoding methods, such as those in the HEVC standard, face inefficiencies in transferring motion information, leading to high bit rates due to the need to transfer accurate motion vectors and associated information, which is not adequately addressed by current merge and non-merge modes.

Innovation Solution

A method that utilizes a candidate predicted motion information list to determine a correspondence between a motion vector difference and a motion vector prediction value, using reference image information and adjacent block motion information to correct the motion vector, thereby reducing the amount of information transferred while maintaining video quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If non-merge mode is used to determine motion information by transferring reference image information, motion vector difference, and motion vector prediction value, then motion information accuracy is improved, but bit rate increases due to large amount of information transferred

Engineering Contradiction:

Improvemotion information accuracyVSAvoidamount of information transferred

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts and separates the motion vector difference from the complete motion information, transmitting only the difference value rather than the full motion vector. This extraction approach reduces the quantity of transmitted data while preserving the ability to reconstruct accurate motion information at the decoder by combining the transmitted difference with predicted motion information.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent introduces motion vector difference as an intermediary element that bridges the reference block motion information and the current block motion information. Instead of directly transmitting complete motion information, the system uses the difference as a mediator that, when added to the predicted motion vector, yields the accurate motion vector for the current block.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Quantity of substance

If merge mode is used to obtain motion vector by completely duplicating adjacent block motion vector, then bit rate is reduced, but motion vector accuracy becomes insufficient

Engineering Contradiction:

Improveinformation transferredVSAvoidmotion vector accuracy

Core Design Contradiction:

Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent changes the parameter representation from complete motion vectors to motion vector differences. By encoding and transmitting only the difference between the predicted motion vector and the actual motion vector, the system achieves more accurate motion information reconstruction while maintaining reduced bit rate compared to traditional merge mode.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12549736B2Bidirectional inter prediction method and terminal based on motion vector difference reduction

Publication Date: 2026.02.10 HUAWEI TECH CO LTD

AI summary

An inter prediction method includes obtaining predicted motion information of a current image block, where the predicted motion information is used to obtain one or more motion vectors of the current image block, and the motion vectors include a forward MV and a backward MV for a bidirectional inter prediction. The disclosed method further includes determining a motion vector difference for the current image block, and calculating the motion vectors based on the predicted motion information and the motion vector difference, where the motion vector difference is shared by the forward MV and the backward MV.