Y-UV Feature Fusion in Image Decompression for Lower Compute

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image/video compression methods using convolutional neural networks face inefficiencies due to separate processing of Y and UV components, leading to high computational complexity and reduced performance.

Innovation Solution

An image decompression method that optimizes the fusion of Y and UV components by performing concatenation and convolution operations, allowing for flexible channel management to reduce computational load while enhancing performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If separate processing of Y and UV components is performed, then processing simplicity is maintained, but computational complexity increases and performance decreases

Engineering Contradiction:
Improveprocessing structureVSAvoidcomputational complexity
Core Design Contradiction:
Device complexityVSPower

Solution Approach 1:

The patent merges the separate processing paths of Y and UV components into a unified processing framework. The encoder and decoder networks are integrated to perform joint optimization, where features from both components are processed together through shared convolutional layers and optimization objectives, reducing overall computational complexity while improving compression performance

Inventive Principle:
Principle #5Merging (Combining)

2Device complexity

If separate processing of Y and UV components is performed, then processing simplicity is maintained, but performance decreases

Engineering Contradiction:
Improveprocessing structureVSAvoidcompression performance
Core Design Contradiction:
Device complexityVSReliability

Solution Approach 1:

The patent combines Y and UV component processing into an integrated neural network framework that performs joint optimization. The encoder network processes both components simultaneously with shared feature extraction, and the decoder reconstructs them through coordinated operations, achieving superior compression performance compared to separate processing

Inventive Principle:
Principle #5Merging (Combining)

3Power

If channel counts are reduced during fusion, then computational power requirements decrease, but information loss may increase

Engineering Contradiction:
Improvecomputational power requirementsVSAvoidfeature information
Core Design Contradiction:
PowerVSLoss of information

Solution Approach 1:

The patent dynamically adjusts the number of channels in feature tensors during the fusion process. The convolutional layers transform feature tensors with different channel counts into unified representations, allowing flexible parameter changes that reduce computational load while preserving essential information through learned feature transformations

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4693195A1Image decompression method and device
Publication Date: 2026.02.11 HUAWEI TECH CO LTD
  • EP4693195A1 patent drawingFigure 1
  • EP4693195A1 patent drawingFigure 2
  • EP4693195A1 patent drawingFigure 3

AI summary

This application provides an image decompression method and apparatus. The image decompression method in this application includes: obtaining a first feature tensor, where the first feature tensor corresponds to a first component of a reconstructed image; obtaining a second feature tensor, where the second feature tensor corresponds to a second component of the reconstructed image; performing concatenation and convolution on the first feature tensor and the second feature tensor to obtain a third feature tensor; obtaining a fourth feature tensor based on the third feature tensor and the second feature tensor; and obtaining the reconstructed image based on the first feature tensor and the fourth feature tensor. In embodiments of this application, a structure for fusing a Y component and a UV component is optimized. This can reduce a quantity of channels, to save computing power and implement lightweight. Even if the quantity of channels is not reduced, performance can still be greatly improved.