In-Loop Reshaping for Video Coding Efficiency

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current video processing technologies face challenges in efficiently managing bandwidth demand due to the increasing use of high-resolution video, particularly in digital communication networks, where existing compression methods do not adequately address the need for improved coding efficiency and latency reduction.

Innovation Solution

The implementation of in-loop reshaping (ILR) techniques for video processing, which involves converting video signals between different domains to optimize compression and decoding efficiency, including methods like luma-dependent chroma residue scaling and adaptive filtering, to enhance coding standards such as HEVC and future video codecs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If existing video compression methods are used, then video can be transmitted over digital networks, but bandwidth demand continues to grow with increasing resolution and connected devices

Engineering Contradiction:
Improvevideo transmission efficiencyVSAvoidbandwidth demand
Core Design Contradiction:
ProductivityVSQuantity of substance

Solution Approach 1:

The patent applies in-loop reshaping that transforms video signals between different domains (first domain and second domain) to optimize compression. By changing the representation parameters of video data through domain transformation, the patent achieves better compression efficiency and reduces bandwidth demand while maintaining video quality

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent introduces domain transformation as an additional processing dimension in the video coding pipeline. By converting video signals between different domains during the encoding and decoding loops, the patent creates new degrees of freedom for optimization that go beyond traditional compression methods

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Productivity

If motion information refinement is performed using samples from a single domain, then the process is simple, but coding efficiency is limited

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

Solution Approach 1:

The patent performs motion information refinement by obtaining samples from both a first domain and a second domain, then combining information from these different domains. This multi-domain approach adds a dimensional aspect to motion refinement that improves coding efficiency beyond single-domain methods

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent creates a composite representation by combining motion information and samples from multiple domains. This composite approach integrates advantages from different domains to achieve superior coding efficiency while managing complexity through structured combination

Inventive Principle:
Principle #40Composite materials

3Manufacturing precision

If in-loop reshaping is implemented with multi-domain sampling, then coding efficiency improves, but computational complexity increases

Engineering Contradiction:
Improvevideo quality precisionVSAvoidprocessing complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent manages complexity by systematically changing parameters during domain transformation. By controlling the transformation parameters and sampling strategies between domains, the patent achieves high video quality precision while keeping processing complexity manageable through parameter optimization

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3906681B1Interactions between in-loop reshaping and inter coding tools
Publication Date: 2024.11.13 DOUYIN VISION CO LTD
  • EP3906681B1 patent drawingFigure 1
  • EP3906681B1 patent drawingFigure 2~3
  • EP3906681B1 patent drawingFigure 4A~4B

AI summary

A method for video processing is provided to include performing, for a conversion between a current video block of a video and a coded representation of the video, a motion information refinement process based on samples in a first domain or a second domain; and performing the conversion based on a result of the motion information refinement process, wherein, during the conversion, the samples are obtained for the current video block from a first prediction block in the first domain using an unrefined motion information, at least a second prediction block is generated in the second domain using a refined motion information used for determining a reconstruction block, and reconstructed samples of the current video block are generated based on the at least the second prediction block.