CNN Filter for Image Coding Artifact Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current video compression techniques face challenges in achieving high compression ratios without sacrificing picture quality, particularly with lossy compression methods that introduce visible spatial compression artifacts.

Innovation Solution

The method employs a neural network-based approach, specifically a modified U-net with skip connections, for image modification by generating a correction image through image down-sampling and filtering, and then combining it with the input image to reduce compression artifacts and enhance visual quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of substance

If lossy image compression is used to achieve high compression rates, then the quantity of data is reduced, but visible spatial compression artifacts are introduced and picture quality deteriorates

Engineering Contradiction:

Improvedata quantityVSAvoidpicture quality

Core Design Contradiction:

Loss of substanceVSManufacturing precision

Solution Approach 1:

A neural network-based intermediary processing stage is introduced between compression and final output. The network takes compressed image data as input and generates a corrected output image, acting as a mediator that compensates for compression artifacts while maintaining the benefits of data reduction

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

Traditional mechanical filtering methods are replaced with a neural network system that uses learned patterns and transformations. The network replaces conventional signal processing mechanisms with an adaptive, data-driven approach that can more effectively restore image quality after compression

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Loss of energy

If traditional compression techniques are applied to reduce data size for transmission, then bandwidth requirements are reduced, but picture quality is compromised due to compression artifacts

Engineering Contradiction:

Improvebandwidth consumptionVSAvoidpicture quality

Core Design Contradiction:

Loss of energyVSManufacturing precision

Solution Approach 1:

The neural network serves as an intermediary restoration stage that processes compressed images to remove artifacts. This allows the system to use aggressive compression for bandwidth efficiency while maintaining quality through the intermediary processing step before final delivery

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system accepts compression artifacts as an inevitable byproduct of bandwidth reduction, then uses the neural network to convert this harmful effect into a benefit by selectively removing artifacts while preserving important image features, ultimately achieving both compression and quality

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

Data Source

PatentUS20230076920A1Global skip connection based convolutional neural network (CNN) filter for image and video coding

Publication Date: 2023.03.09 HUAWEI TECH CO LTD

AI summary

The present disclosure relates to image processing and in particular to modification of an image using a processing such as neural network. The processing is performed to generate a correction image based on an input image. Then, the input image is modified by combining it with the correction image. The processing with the neural network includes at least one stage including image down-sampling and filtering of the down-sampled image; and at least one stage of image up-sampling. An advantage of such approach is increased efficiency of the neural network, which may lead to faster learning and improved performance. The embodiments provide methods and apparatuses for the processing with a trained neural network, as well as methods and apparatuses for training of such neural network for image modification.