Merge Estimation Region Partitioning for Parallel Video Encoding

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video encoding and decoding technologies face inefficiencies in parallel merge candidate list construction due to dependencies between neighboring blocks, particularly when using spatial merge candidates, which limits parallel processing and increases computational complexity.

Innovation Solution

The implementation of merge estimation regions (MERs) that constrain partitioning such that each MER contains complete coding units and each coding unit contains complete MERs, enabling parallel merge candidate list construction across these regions, thereby facilitating faster inter prediction processing.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If spatial merge candidates are used for motion vector prediction, then prediction accuracy is improved, but parallel processing capability deteriorates due to dependencies between neighboring blocks

Engineering Contradiction:

Improveprediction accuracyVSAvoidparallel processing capability

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The picture is divided into multiple merge estimation regions (MERs), where each MER is a rectangular region that is independent for merge candidate list construction. This segmentation allows parallel processing across different MERs while maintaining spatial merge candidate accuracy within each region, as the dependency is confined within MER boundaries rather than across the entire picture.

Inventive Principle:

Principle #1Segmentation

2Productivity

If merge estimation regions are introduced to enable parallel processing, then processing speed is improved, but device complexity increases due to additional partitioning constraints

Engineering Contradiction:

Improveprocessing speedVSAvoidpartitioning constraint complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The MER size is configured as a parameter that can be adjusted to balance between parallel processing efficiency and constraint complexity. By changing the MER size parameter, the system can optimize the trade-off between enabling sufficient parallelism and maintaining manageable partitioning constraints, without requiring complex adaptive algorithms.

Inventive Principle:

Principle #35Parameter changes

3Productivity

If coding units are constrained to be completely within MERs, then parallel merge candidate list construction is enabled, but partitioning flexibility deteriorates

Engineering Contradiction:

Improveparallel merge candidate list constructionVSAvoidpartitioning flexibility

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The partitioning structure is made dynamic by allowing the video coder to adaptively choose between different partitioning modes (binary tree split, ternary tree split, or no split) based on the MER configuration and content characteristics. This dynamic adaptability maintains partitioning flexibility while ensuring that final coding unit boundaries align with MER boundaries to enable parallel processing.

Inventive Principle:

Principle #15Dynamics

Data Source

PatentUS11539982B2Merge estimation region for multi-type-tree block structure

Publication Date: 2022.12.27 QUALCOMM INC

AI summary

A video encoder may encode a picture of video data using merge estimation regions (MERs). The video encoder may determine merge candidate lists in parallel for coding units within a MER. The video encoder may also partition the picture of video data into coding units according to a constraint, wherein the constraint specifies that the partitioning is constrained such that, for each MER containing one or more coding units, the one or more coding units are completely in the MER, and for each coding unit containing one or more MERs, the MERs are completely in the coding unit.