Chroma Subsampling Handling for ML Picture Coding Trade-Offs

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing video coding technologies using machine learning networks face inefficiencies in handling chroma subsampled formats, leading to challenges in bitrate and quality trade-offs, particularly in applications with limited network resources and high video quality demands.

Innovation Solution

Adaptive selection between luma down-sampling and chroma up-sampling to match resolution, enabling efficient encoding and decoding of picture portions using variational autoencoders, with adaptive switching based on cost functions and trained networks to optimize bitrate and quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If chroma subsampled formats are used to reduce bitrate, then data transmission efficiency improves, but picture quality deteriorates due to loss of chroma information

Engineering Contradiction:

ImprovebitrateVSAvoidpicture quality

Core Design Contradiction:

Quantity of substanceVSManufacturing precision

Solution Approach 1:

The patent applies dynamics by making the chroma resolution adaptable rather than fixed. The system dynamically adjusts chroma component resolution based on content characteristics and encoding parameters, allowing the chroma subsampling level to vary across different regions and scenes. This enables optimal balance between bitrate efficiency and picture quality for different video content types.

Inventive Principle:

Principle #15Dynamics

Solution Approach 2:

The patent implements local quality by applying different chroma resolution levels to different spatial regions of the picture. Important regions with significant chroma information maintain higher resolution, while less important regions use lower resolution subsampling. This localized approach preserves picture quality where needed while reducing bitrate in areas where chroma detail is less critical.

Inventive Principle:

Principle #3Local quality

2Manufacturing precision

If machine learning networks are used to improve picture quality, then decoding complexity increases, but processing speed decreases

Engineering Contradiction:

Improvepicture qualityVSAvoidprocessing speed

Core Design Contradiction:

Manufacturing precisionVSProductivity

Solution Approach 1:

The patent applies preliminary action by pre-training the machine learning network during an offline phase. The neural network is trained on large datasets to learn optimal chroma reconstruction patterns, and this training is completed beforehand. During actual encoding and decoding operations, the pre-trained network is applied directly without requiring real-time training, significantly improving processing speed while maintaining picture quality benefits.

Inventive Principle:

Principle #10Preliminary action

3Device complexity

If luma down-sampling is applied to match chroma resolution, then encoding complexity reduces, but luma information loss increases

Engineering Contradiction:

Improveencoding complexityVSAvoidluma information

Core Design Contradiction:

Device complexityVSLoss of information

Solution Approach 1:

The patent applies inversion by reversing the traditional approach. Instead of down-sampling luma to match chroma resolution, the system up-samples chroma to match luma resolution. This allows the full-resolution luma component to be preserved without information loss, while chroma is enhanced to the same resolution level, achieving both simplified encoding and preservation of luma quality.

Inventive Principle:

Principle #13The other way round (Inversion)

4Manufacturing precision

If chroma up-sampling is applied to match luma resolution, then picture quality improves, but bitrate increases

Engineering Contradiction:

Improvepicture qualityVSAvoidbitrate

Core Design Contradiction:

Manufacturing precisionVSQuantity of substance

Solution Approach 1:

The patent applies parameter changes by dynamically adjusting the chroma up-sampling level based on content characteristics. Rather than always up-sampling chroma to full luma resolution, the system modifies the chroma resolution parameter adaptively according to scene complexity, motion levels, and importance of chroma information in different regions. This enables picture quality improvement where needed while controlling overall bitrate increase.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS12513297B2Method for chroma subsampled formats handling in machine-learning-based picture coding

Publication Date: 2025.12.30 HUAWEI TECH CO LTD

AI summary

The present disclosure relates to picture or video encoding and decoding. A picture portion includes a luma component and a chroma component with a resolution lower than a resolution of the luma component. To obtain matching resolution of the luma component and the chroma component, a selection is performed between a down-sampling processing of the luma component and an up-sampling processing of the chroma component. Then, the selected processing is applied to a picture portion. The bitstream is then generated based on the selected processing, and includes encoding the processed picture portion. The selection between chroma up-sampling and luma down-sampling enables adaption of the input picture portion for the desired application. For example, some applications may aim at reducing complexity or bitrate in which case the luma down-sampling may be more favorable, whereas other application may aim at higher quality in which case chroma up-sampling may be more favorable.