Chinese character font generation method and system based on hypergraph neural network and diffusion model

CN121209706BActive Publication Date: 2026-07-03HAOGE DESIGN (SHENZHEN) CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
HAOGE DESIGN (SHENZHEN) CO LTD
Filing Date
2025-09-26
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing generative diffusion models fail to effectively consider the differences between font images and natural images in Chinese font generation, resulting in generated fonts that lose details, have inconsistent styles, low resolution, and low running efficiency.

Method used

By employing a hypergraph neural network and a diffusion model, a hypergraph structure is constructed by extracting the radicals, components, strokes, and structural information of characters. This structure is then combined with the hypergraph neural network and a conditional generation model to guide font generation, achieving style consistency and high resolution.

Benefits of technology

The generated Chinese fonts have good style consistency and high resolution, accurately expressing the target style. This solves the problems of inconsistency and low resolution in existing font generation technologies, and improves generation efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121209706B_ABST
    Figure CN121209706B_ABST
Patent Text Reader

Abstract

The application provides a Chinese font generation method and system based on a hypergraph neural network and a diffusion model, extracts example character information features through a structure information generator, encodes the example characters through an encoder, takes the example character encoding features as nodes, connects the nodes sharing any same character information feature into a hyperedge, thereby constructing a hypergraph structure, then inputs the hypergraph neural network to obtain style encoding features of the example characters, extracts encoding features of to-be-generated characters at the same time, and inputs the encoding features into a conditional generation model to guide font generation of the diffusion model. Since the style features of multiple example characters are fused, the target style character generated by the to-be-generated character can accurately represent the style of the example character using a small amount of example characters.
Need to check novelty before this filing date? Find Prior Art