Dynamic Image Encoding Motion Vector Prediction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The existing moving picture encoding and decoding technologies, such as MPEG-4 AVC/H.264, face challenges in maintaining prediction accuracy and encoding efficiency, particularly in the time direct mode.

Innovation Solution

The proposed solution involves deriving multiple motion vector predictors from adjacent blocks and calculating the difference vector between the target block's motion vector and the selected predictor, which is then encoded to reduce the overall code amount.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If motion compensation prediction is used with multiple reference pictures, then prediction accuracy is improved, but coding information amount increases

Engineering Contradiction:
Improveprediction accuracyVSAvoidcoding information amount
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent derives multiple motion vector predictors in advance from adjacent blocks before encoding the target block. By preparing these predictors beforehand, the system can select the most appropriate one without needing to encode multiple motion vectors, thus maintaining prediction accuracy while reducing coding information.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent changes the approach from encoding multiple motion vectors to encoding a difference value between the target motion vector and the selected predictor. This parameter transformation reduces the number of bits required to represent motion information while preserving prediction accuracy.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If multiple motion vectors are encoded for each block, then prediction accuracy is improved, but encoding complexity increases

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

Solution Approach 1:

The patent extracts only the essential motion information by deriving predictors from adjacent blocks and selecting the best match. Instead of encoding all possible motion vectors, the system extracts and encodes only the difference between the target vector and the selected predictor, simplifying the encoding process.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

Motion vector predictors are derived in advance from neighboring blocks before the actual encoding of the target block. This preliminary derivation reduces the computational complexity during the main encoding process by eliminating the need to calculate and compare multiple motion vectors from scratch.

Inventive Principle:
Principle #10Preliminary action

3Loss of information

If difference vector encoding is used, then code amount is reduced, but prediction accuracy may deteriorate

Engineering Contradiction:
Improvecode amountVSAvoidprediction accuracy
Core Design Contradiction:
Loss of informationVSMeasurement precision

Solution Approach 1:

The patent uses a feedback mechanism where multiple motion vector predictors are derived from adjacent blocks and the best predictor is selected based on matching quality. This feedback loop ensures that the difference vector encoding maintains prediction accuracy by adaptively selecting the most appropriate predictor for each block.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent transforms the motion vector representation from absolute values to difference values relative to selected predictors. By changing the parameter representation and using adaptive predictor selection, the system reduces code amount while preserving prediction accuracy through intelligent parameter transformation.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3907997B1Dynamic image encoding device, dynamic image encoding method, dynamic image encoding program, dynamic image decoding device, dynamic image decoding method, and dynamic image decoding program
Publication Date: 2025.04.09 JVC KENWOOD CORP
  • EP3907997B1 patent drawingFigure 1
  • EP3907997B1 patent drawingFigure 2
  • EP3907997B1 patent drawingFigure 3

AI summary

During inter-prediction, whereby a first block having each picture divided therein is divided into one or a plurality of second blocks, a spatial merge candidate generation unit (130) derives spatial merge candidates, without referring to blocks included in the first block, including the second block. A reference index derivation unit (131) for temporal merge candidates: sets reference index information for a temporal merge candidate as the value for reference index information for an encoded prediction block approaching the left side of a prediction block for encoding, when in a mode whereby the encoded block is divided into upper and lower prediction blocks at the horizontal interface; and sets reference index information for a temporal merge candidate as the value for reference index information for an encoded prediction block approaching the top side of the prediction block for encoding, when in a mode whereby the encoded block is divided into left and right prediction blocks at the vertical interface.