<b>EdgeStyleGAN: An Optimization Approach for Chinese Character Font Generation Based on Font Contours</b>

With the rapid development of the character font industry, utilizing deep learning image generation models to assist in developing Chinese character fonts has recently become a research focus and developmental trend. Presently, methods based on deep generative models, particularly those employing generative adversarial networks (GANs), dominate the field. However, these models exhibit limitations in character font development, such as the inability to depict font contours clearly and the excessive training samples requirement. In response to these challenges, this paper proposes a Chinese character font generation method based on font contour information (EdgeStyleGAN). This approach introduces an edge extraction module into the style transfer network utilizing convolutional operations to extract the edge contour information from the source style font images, which is subsequently conveyed to the generator. The generator concatenates these edge details with the source style font images to better comprehend the font's structure and edge intricacies, enhancing Chinese character image generation quality. Additionally, a training dataset comprising 500 Chinese characters was constructed; this dataset encompasses most character structures and helps designers swiftly establish design styles and rules for font libraries, ensuring that the model accurately captures the stylistic nuances of the target font. The experimental results demonstrate that the proposed method successfully accomplishes generation tasks with commonly used styles of Chinese character fonts. Its effectiveness is validated through qualitative and quantitative assessments. This research provides robust support for applying intelligently generated Chinese characters in the character font industry.