Multi-Layer Inter Coding with Conditional Motion Refinement

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video coding technologies face challenges in handling motion vector refinement when reference pictures are from a different layer than the current picture, leading to errors such as division by zero, which degrade the user experience.

Innovation Solution

The ability to selectively disable motion vector refinement when reference pictures are from a different layer than the current picture, using a motion vector refinement flag to manage this process, and enabling reference picture resampling when necessary.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If motion vector refinement is applied when reference pictures are from a different layer, then coding precision is improved, but coding errors occur (division by zero)

Engineering Contradiction:

Improvecoding precisionVSAvoidcoding error avoidance

Core Design Contradiction:

Manufacturing precisionVSReliability

Solution Approach 1:

The patent changes the state of the motion vector refinement process by introducing a conditional parameter: when reference pictures are from a different layer, the motion vector refinement is disabled. This parameter change (enabling/disabling based on layer condition) resolves the contradiction by preventing coding errors while maintaining precision when appropriate

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent makes the motion vector refinement process dynamic by allowing it to be enabled or disabled based on the layer relationship between reference pictures and current picture. This dynamic adaptation resolves the contradiction by adjusting the refinement process according to specific coding conditions

Inventive Principle:

Principle #15Dynamics

2Reliability

If motion vector refinement is disabled for cross-layer references, then coding errors are avoided, but compression efficiency decreases

Engineering Contradiction:

Improvecoding error avoidanceVSAvoidcompression efficiency

Core Design Contradiction:

ReliabilityVSProductivity

Solution Approach 1:

The patent applies local quality by treating different layer relationships differently: motion vector refinement is disabled only for cross-layer references (local condition) while potentially remaining enabled for same-layer references. This localized approach maintains reliability where needed while preserving compression efficiency elsewhere

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The patent segments the motion vector refinement process into conditional paths: one path for same-layer references and another for cross-layer references. This segmentation allows the system to maintain high compression efficiency for common same-layer cases while ensuring reliability for cross-layer cases

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20250301159A1Inter Coding in Video Coding with the Support of Multiple Layers

Publication Date: 2025.09.25 HUAWEI TECH CO LTD

AI summary

A non-transitory computer-readable storage medium storing instructions that, when executed by one or more processors, cause a video processing apparatus to: receive a coded video bitstream containing reference picture lists; obtain reference pictures for a current picture from the reference picture lists; set a decoder motion vector refinement (DMVR) flag to a first value to enable DMVR for a current block of the current picture when the reference pictures are in a same layer as the current picture; set the DMVR flag to a second value to disable the DMVR for the current block of the current picture when the reference pictures are in a different layer than the current picture; and refine a motion vector corresponding to the current block when the DMVR flag is set to the first value.