Font Style Fusion Model for Chinese Character Structure Preservation

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing image translation technologies are ineffective in modifying structural information of Chinese characters, leading to issues such as broken strokes, uneven edges, and lost or redundant strokes in font data style transfer, resulting in font fusion outcomes that differ significantly from actual requirements.

Innovation Solution

A method and apparatus for generating a word by obtaining images corresponding to a word to be processed and a reference word, and inputting these images into a target font style fusion model to obtain a target word with a font style between the two input font styles.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Shape

If style transfer or image translation technology is used to modify font styles, then texture modification is improved, but structural information is lost resulting in broken strokes, uneven edges, and lost or redundant strokes

Engineering Contradiction:

Improvefont textureVSAvoidstroke structure

Core Design Contradiction:

ShapeVSManufacturing precision

Solution Approach 1:

The patent segments the font representation into two independent parts: structural information (stroke topology, edge connectivity) and textural information (surface characteristics, shading). This segmentation allows the structural component to be preserved while the textural component is transferred from the reference font, preventing the breakdown of stroke structures that occurs in conventional style transfer methods.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an intermediary representation layer that decouples structure and texture. This intermediary structure preserves the topological and geometric relationships of strokes while allowing texture mapping from the reference font. The intermediary mechanism ensures that structural integrity is maintained during the style transfer process.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Adaptability or versatility

If conventional font data style transfer is applied, then font style modification is achieved, but the generated font fusion results differ significantly from actual requirements

Engineering Contradiction:

Improvefont style variationVSAvoidfont fusion accuracy

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The patent applies local quality by differentiating between structural regions and textural regions. The structural representation maintains consistent stroke connectivity and topological properties locally, while the textural representation captures local surface characteristics. This localized differentiation ensures that each region retains its appropriate properties, improving the reliability of font fusion results.

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

Instead of transferring only texture over a fixed structure, the patent inverts the approach by using the structure to guide and constrain the texture transfer. The structural information determines where and how texture is applied, ensuring that the final font fusion accurately reflects the intended style while maintaining structural correctness.

Inventive Principle:

Principle #13The other way round (Inversion)

Data Source

PatentUS20250061623A1Word generation method, and apparatus, electronic device, and storage medium

Publication Date: 2025.02.20 BEIJING ZITIAO NETWORK TECH CO LTD

AI summary

The disclosure provides a method and apparatus for generating a word, an electronic device, and a storage medium. The method for generating a word includes: obtaining images corresponding to a word to be processed and a reference word; and inputting the image to be processed into a target font style fusion model to obtain a target word of the word to be processed in a target font style. The target font style is determined based on the target font style fusion model fusing a reference font style of the reference word and a font style to be processed of the word to be processed.