Semantic Image Encoding for High-Compression Data Reduction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image compression algorithms, such as JPEG, have limited compression rates and result in significant data volume post-compression, necessitating more efficient data reduction methods.

Innovation Solution

Divide an input image into multiple areas using semantic segmentation, convert each area into semantic text data using AI, and package these data to generate encoded text data, reducing data volume.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional image compression algorithms (JPEG) are used to remove low-frequency redundant data, then processing stability is improved, but compression rate deteriorates and data volume remains large

Engineering Contradiction:

Improvealgorithm stabilityVSAvoiddata volume

Core Design Contradiction:

ReliabilityVSQuantity of substance

Solution Approach 1:

The input image is divided into multiple image areas according to a preset division mode before conversion. This segmentation allows the system to process and compress different regions independently, enabling more effective data reduction while maintaining important visual information in each segment

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent replaces traditional mechanical image compression algorithms with an AI-based semantic conversion system. Instead of using mathematical transforms to remove redundant data, the system converts image areas into semantic text data using artificial intelligence, achieving superior compression rates while preserving meaningful information

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Quantity of substance

If image data is converted to semantic text data using AI conversion modules, then compression rate is improved and data volume is reduced, but processing complexity increases

Engineering Contradiction:

Improvedata volumeVSAvoidconversion system complexity

Core Design Contradiction:

Quantity of substanceVSDevice complexity

Solution Approach 1:

The conversion system is divided into multiple specialized conversion modules (first conversion module, second conversion module, third conversion module), each handling specific image areas or conversion tasks. This modular architecture manages complexity by breaking down the overall conversion process into smaller, more manageable components

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces packaged semantic text data as an intermediary representation between the original image and the final decoded output. This intermediate format serves as a compact bridge, reducing data volume while maintaining the ability to reconstruct the image, thus managing the complexity of AI-based conversion

Inventive Principle:

Principle #24Intermediary (Mediator)

3Quantity of substance

If semantic text data is used to represent image areas, then storage space is reduced, but information loss may occur during conversion

Engineering Contradiction:

Improvestorage spaceVSAvoidimage information loss

Core Design Contradiction:

Quantity of substanceVSLoss of information

Solution Approach 1:

The system employs multiple conversion modules with different conversion parameters and strategies. By adjusting conversion parameters and selecting appropriate modules based on image characteristics, the system optimizes the balance between compression ratio and information preservation, minimizing information loss while reducing storage requirements

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20260004463A1Image encoding method and decoding method

Publication Date: 2026.01.01 LENOVO (BEIJING) LTD

AI summary

An image encoding method includes dividing an input image into a plurality of image areas according to a preset division mode, converting the plurality of image areas into a plurality of pieces of semantic text data based on a first conversion module, and packaging the plurality of pieces of semantic text data to generate encoded data of the input image. Each piece of semantic text data represents semantics describing a corresponding image area.