Parallel Merge Candidate Generation for Video Coding Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video coding technologies face inefficiencies in determining merge candidates for motion vector prediction, particularly in the HEVC standard, due to the need for comparing motion information across prediction units within a coding unit, which increases computational complexity and limits parallel processing.

Innovation Solution

The proposed technique eliminates the need for comparing motion information between candidate blocks and other prediction units by generating merge candidate sets without considering the prediction unit's index, allowing for parallel processing and removing redundant candidates located within other prediction units, thus simplifying the merge candidate generation process.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If motion information comparison is performed across prediction units within a coding unit to determine merge candidates, then accurate inter-prediction quality is maintained, but computational complexity increases and parallel processing is limited

Engineering Contradiction:
Improveinter-prediction qualityVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent divides the coding unit into multiple prediction units and generates merge candidate sets independently for each prediction unit without requiring comparison across prediction unit boundaries. This segmentation enables parallel processing while maintaining adequate prediction quality within each unit.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent removes the requirement for cross-prediction-unit motion information comparison from the merge candidate generation process. By extracting this comparison step, the patent reduces computational complexity and enables independent parallel generation of merge candidates for each prediction unit.

Inventive Principle:
Principle #2Taking out (Extraction)

2Reliability

If motion information comparison is performed across prediction units within a coding unit to determine merge candidates, then accurate inter-prediction quality is maintained, but parallel processing capability is reduced

Engineering Contradiction:
Improveinter-prediction qualityVSAvoidparallel processing capability
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent segments the merge candidate generation process into independent operations for each prediction unit, allowing simultaneous execution across multiple prediction units without inter-dependent comparisons, thereby enabling full parallel processing capability.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs merge candidate generation for each prediction unit independently and simultaneously, rather than sequentially comparing motion information across units. This preliminary independent generation enables parallel processing while maintaining adequate prediction quality.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If merge candidates are generated by considering prediction unit index and comparing motion information, then accurate motion prediction is achieved, but redundant candidate signaling occurs

Engineering Contradiction:
Improvemotion prediction accuracyVSAvoidredundant signaling
Core Design Contradiction:
ReliabilityVSLoss of substance

Solution Approach 1:

The patent removes the prediction unit index consideration and cross-unit motion information comparison from the merge candidate generation process, eliminating redundant candidate signaling while maintaining adequate motion prediction accuracy within each prediction unit.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent discards the requirement for cross-prediction-unit motion information comparison and prediction unit index tracking, eliminating redundant signaling overhead while recovering adequate motion prediction performance through independent per-unit candidate generation.

Inventive Principle:
Principle #34Discarding and recovering

Data Source

PatentEP2721821B1Parallelization friendly merge candidates for video coding
Publication Date: 2020.11.25 QUALCOMM INC
  • EP2721821B1 patent drawingFigure 1
  • EP2721821B1 patent drawingFigure 2
  • EP2721821B1 patent drawingFigure 3A~3B

AI summary

This disclosure presents methods and systems for coding video in merge mode of a motion vector prediction process. A method of coding video data may determining a merge candidate set for a current prediction unit of a current coding unit, wherein the merge candidate set is determined without comparing motion information of a merge candidate in the merge candidate set to motion information of any other prediction units, and performing a merge motion vector prediction process for the current prediction unit using the merge candidate set. The method may further comprise excluding merge candidates from the merge candidate set that are within another prediction unit of the current coding unit.