Multi-view video motion vector prediction using inter-view references

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current multi-view video encoding and decoding techniques face inefficiencies due to the increasing data amount with the number of views, necessitating an effective method for motion vector and residual prediction to enhance compression efficiency.

Innovation Solution

A method for encoding and decoding multi-view videos through motion vector prediction, which determines suitable prediction schemes for current and co-located blocks, generating motion vector predictors using long-term, short-term, or inter-view predictions, and performing residual prediction by calculating prediction residuals based on temporal or inter-view references.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If multi-view video encoding is performed with increasing number of views, then video coverage and viewing experience are improved, but data amount and compression complexity increase significantly

Engineering Contradiction:
Improvemulti-view video coverageVSAvoiddata amount
Core Design Contradiction:
Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent merges motion vector prediction across multiple views by establishing inter-view prediction relationships. Motion vectors from reference views are combined with temporal motion vectors to generate predictors for current view blocks, effectively merging spatial and temporal prediction information to reduce redundant data representation across views

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent uses motion vectors from co-located blocks in reference views as copied predictors for the current view. By copying and adapting motion information from temporally and spatially corresponding blocks across views, the system reduces the need to encode complete motion information independently for each view

Inventive Principle:
Principle #26Copying

2Productivity

If conventional motion vector prediction is used for multi-view video, then encoding simplicity is maintained, but compression efficiency deteriorates due to redundancy

Engineering Contradiction:
Improvecompression efficiencyVSAvoiddata redundancy
Core Design Contradiction:
ProductivityVSLoss of information

Solution Approach 1:

The patent implements feedback mechanisms where motion vector predictors are adaptively selected based on prediction mode flags and reference picture availability. The system uses feedback from prediction mode decisions to determine whether to use inter-view temporal motion vector prediction, inter-view spatial prediction, or conventional temporal prediction, optimizing compression by selecting the most efficient predictor for each block

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent introduces dynamic selection of motion vector prediction modes based on block characteristics and reference picture availability. The system dynamically switches between different prediction strategies (inter-view temporal, inter-view spatial, conventional temporal) depending on the specific coding context, allowing adaptive optimization of compression efficiency across different video content and view configurations

Inventive Principle:
Principle #15Dynamics

3Productivity

If inter-view prediction is applied to all blocks, then compression efficiency is improved, but device complexity and processing overhead increase

Engineering Contradiction:
Improvecompression efficiencyVSAvoidprocessing complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent applies local quality by enabling inter-view prediction only for blocks where reference pictures are available and prediction modes are appropriate, rather than uniformly applying it to all blocks. The system locally adapts the prediction strategy based on the specific block's characteristics, reference picture list configuration, and motion characteristics, reducing unnecessary processing complexity in regions where inter-view prediction is not beneficial

Inventive Principle:
Principle #3Local quality

Data Source

PatentEP3958572B1Method for encoding multi-view video, method for decoding multi-view video and recording medium therefore
Publication Date: 2024.02.07 DOLBY LABORATORIES LICENSING CORP
  • EP3958572B1 patent drawingFigure 1~2
  • EP3958572B1 patent drawingFigure 3~4
  • EP3958572B1 patent drawingFigure 5

AI summary

Disclosed is a technique related to a method for a motion vector prediction and a residual prediction for a multi-view video and an apparatus for performing the method. A method for decoding a motion vector for a multi-view video comprises the steps of: determining a motion prediction method performed on a current block which is an object to be decoded and a corresponding block corresponding to the current block; and generating a motion vector prediction value of the current block using a motion vector of the corresponding block on the basis of the determined motion prediction method. Thus, a temporal motion vector can be adaptively predicted according to a motion vector prediction method of the current block and the corresponding block.