Handwritten fonts possess a certain level of affinity,particularly in book layout design,exhibiting humanization and expressiveness.However,manual font design processes are cumbersome and demand high levels of expertise.An effective approach to address this issue is utilizing artificial intelligence-based handwritten font generation algorithms to aid in design.In this paper,a few-shot font generation network was employed,wherein multiple style features were extracted using multi-sub-encoders for better capturing diverse local concepts.In terms of sample training,an end-to-end training approach was adopted,significantly reducing training time and enhancing font generation efficiency.Through a combination of qualitative and quantitative analysis,the various font generation methods were extensively compared.Compared to other methods,the approach proposed in this paper yielded lower FID and LPIPS values,adequately demonstrating its superiority in generation performance,with generated fonts being clearer,meeting the requirement of high fidelity.This method provides a faster and more effective solution for book layout design,streamlining the cumbersome process of font design and enhancing design efficiency.Future research can further optimize the quality and diversity of generated fonts to meet the needs of different book layout designs.
NETL
NSTL
万方数据
