Implicit Adaptive Motion Vector Predictor Selection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing video encoding standards, such as MPEG-4 AVC, face challenges in predicting motion vectors effectively, especially in complex motion fields, leading to quality degradation and increased complexity and overhead in motion vector predictor selection, particularly in low bitrate applications.

Innovation Solution

The implementation of implicit adaptive motion vector predictor selection, which defines a set of candidates based on spatio-temporal neighboring motion vectors and selects the best predictor based on their fidelity, without explicit signaling, leveraging correlation among neighboring vectors to achieve accurate prediction without additional overhead.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If the median of spatial causal neighboring motion vectors is used for prediction, then the method is simple and effective for smooth motion fields, but the quality degrades when the motion field is complicated or neighboring vectors have low fidelity

Engineering Contradiction:
Improveprediction method complexityVSAvoidmotion vector prediction quality
Core Design Contradiction:
Device complexityVSManufacturing precision

Solution Approach 1:

The patent applies dynamics by making the prediction method adaptive rather than static. The system dynamically selects between different prediction strategies (spatial, temporal, or combined) based on the characteristics of the current block and neighboring vectors. This allows the method to adapt to different motion field complexities, maintaining simplicity for smooth regions while improving accuracy for complex regions without uniformly increasing complexity across all blocks.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes parameters by introducing fidelity metrics and block type classifications that modify how motion vectors are selected and weighted. Instead of always using the simple median, the system adjusts prediction parameters based on neighboring vector fidelity values and block characteristics, enabling quality improvement in complex regions while maintaining efficiency elsewhere through parameter-based adaptation.

Inventive Principle:
Principle #35Parameter changes

2Manufacturing precision

If motion vector competition (MVComp) is used to explicitly transmit the best motion vector predictor index, then the prediction accuracy is improved through rate-distortion optimization, but the overhead and complexity increase significantly

Engineering Contradiction:
Improvemotion vector prediction accuracyVSAvoidbitrate overhead
Core Design Contradiction:
Manufacturing precisionVSLoss of information

Solution Approach 1:

The patent extracts only the essential information needed for improved prediction without transmitting full predictor indices. Instead of sending explicit index information for multiple candidate predictors as in MVComp, the system extracts and transmits only fidelity metrics and block type indicators, allowing the decoder to reconstruct the same prediction decisions locally. This reduces overhead while maintaining prediction accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent enables self-service by providing the decoder with sufficient information (fidelity metrics and block types) to independently determine the best motion vector predictor without requiring explicit index transmission. Both encoder and decoder use the same fidelity calculations and selection logic, allowing the system to serve itself by eliminating the need for additional predictor index signaling while achieving rate-distortion optimized predictions.

Inventive Principle:
Principle #25Self-service

3Manufacturing precision

If motion vector competition (MVComp) checks and compares each motion vector predictor in the mode decision loop, then the best predictor is selected through rate-distortion optimization, but the encoder complexity becomes very high

Engineering Contradiction:
Improvemotion vector predictor selection accuracyVSAvoidencoder complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by calculating fidelity metrics for all candidate motion vectors before the mode decision loop. These pre-computed fidelity values and block type classifications are stored and reused during mode decision, eliminating the need to re-evaluate predictor quality for each mode. This preliminary preparation reduces encoder complexity while maintaining the ability to select the best predictor through rate-distortion optimization.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent uses partial action by selectively applying complex fidelity-based selection only where needed rather than uniformly across all blocks. The system performs detailed predictor evaluation only for blocks where neighboring vectors show varying fidelity or where block types indicate complex motion, while using simpler methods for straightforward cases. This partial application of complex processing reduces overall encoder complexity while maintaining prediction accuracy where it matters most.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2548372B1Methods and apparatus for implicit adaptive motion vector predictor selection for video encoding and decoding
Publication Date: 2023.02.15 INTERDIGITAL MADISON PATENT HLDG
  • EP2548372B1 patent drawingFigure 1
  • EP2548372B1 patent drawingFigure 2
  • EP2548372B1 patent drawingFigure 3

AI summary

Methods and apparatus are provided for implicit adaptive motion vector predictor selection for video encoding and decoding. The method encodes an image block using implicit adaptive motion vector predictor selection. The motion vector for encoding the block is predicted by defining (315) a set of motion vector predictor candidates for the block based on a plurality of spatio-temporal neighboring motion vectors; and then selecting (325) from the set a particular motion vector predictor candidate as the motion vector for encoding the bloc. The selection is made responsive to respective values for a fidelity of the spatio-temporal neighboring motion vectors (320).