HMVP Motion Vector Prediction for Parallel Tile Video Coding

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video coding technologies face challenges in achieving efficient compression of video data without sacrificing picture quality, particularly in limited bandwidth and memory resource scenarios.

Innovation Solution

Implementing history-based motion vector prediction (HMVP) for coding and decoding processes, including initializing and managing a History Based Motion Vector Prediction (HMVP) list for Coding Tree Units (CTUs) within tiles, allowing for parallel processing and efficient data representation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of substance

If video data is compressed using conventional methods, then the data size is reduced for transmission, but the compression ratio is insufficient and picture quality deteriorates

Engineering Contradiction:

Improvevideo data sizeVSAvoidpicture quality

Core Design Contradiction:

Loss of substanceVSManufacturing precision

Solution Approach 1:

The patent applies preliminary action by initializing the HMVP list at the beginning of each tile or tile group before processing CTUs. This pre-initialization ensures that motion vector prediction candidates are prepared in advance, enabling efficient compression without compromising picture quality during the actual encoding process

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The patent segments the video picture into tiles and tile groups, with each having independent HMVP list initialization. This segmentation allows parallel processing of different tiles while maintaining compression efficiency and picture quality through localized motion vector prediction

Inventive Principle:

Principle #1Segmentation

2Productivity

If parallel processing is implemented for CTUs, then processing efficiency is improved, but data dependency between CTUs must be managed

Engineering Contradiction:

Improveprocessing efficiencyVSAvoiddata dependency management

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent divides the picture into multiple tiles and tile groups that can be processed in parallel. Each tile group has independent HMVP list initialization, eliminating data dependencies between tiles and enabling simultaneous processing while maintaining processing efficiency

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent applies local quality by initializing HMVP lists independently for each tile or tile group based on local characteristics. This allows each region to be processed with appropriate motion vector prediction candidates while enabling parallel processing without complex inter-tile data dependency management

Inventive Principle:

Principle #3Local quality

Data Source

PatentUS20250343895A1Encoder, decoder and corresponding methods using history based motion vector prediction

Publication Date: 2025.11.06 HUAWEI TECH CO LTD

AI summary

Embodiments provide methods and devices (encoder and/or decoder) of coding a picture. A History Based Motion Vector Prediction (HMVP) list for a current Coding Tree Unit (CTU) row within a tile of a picture is initialized and a CTU of the current CTU row is processed based on the initialized HMVP list.