Citation
Abstract
The process of designing fonts traditionally requires a great deal of manpower, taking up to one year to complete one style set. In computer vision (CV) and computer graphics (CG), the automatic generation of Chinese fonts with a large number of complex glyphs remains a challenging and persistent problem. In this article, we propose an end-to-end network for generating 9169 Chinese characters without the need for human intervention. Due to the similarity of the strokes of the Chinese characters, different strokes will be recognized as the same type of stroke, resulting in errors. In this article maps the semantic information of Chinese character stroke categories to different stages of the encoder and adjusts the stroke category semantics in specific channels to ensure the synthesis of correct strokes. A deformation attention Skip-connection module was designed to adapt to font generation tasks by learning offsets in features from the decoder and encoder, weights in a cross-channel interactive manner, and adaptively re-scaling features using the learned weights and offsets. There are only a few parameters involved in this method, but it provides significant performance improvements. Compared to the prior art, the method employed by us outperforms it in terms of generating commercial fonts and handwritten fonts, which can help font designers with font design and improve the efficiency of font libraries. © 2023 Elsevier Ltd
Download File
Full text not available from this repository.
Official URL or Download Paper: https://www.scopus.com/inward/record.uri?eid=2-s2....
|
Additional Metadata
Item Type: | Article |
---|---|
Divisions: | Faculty of Computer Science and Information Technology |
DOI Number: | https://doi.org/10.1016/j.eswa.2023.121407 |
Publisher: | Elsevier |
Keywords: | Deformable convolution; Stroke semantics; Generative models; Chinese font generation |
Depositing User: | Ms. Nuraida Ibrahim |
Date Deposited: | 13 Feb 2024 04:10 |
Last Modified: | 13 Feb 2024 04:10 |
Altmetrics: | http://www.altmetric.com/details.php?domain=psasir.upm.edu.my&doi=10.1016/j.eswa.2023.121407 |
URI: | http://psasir.upm.edu.my/id/eprint/105692 |
Statistic Details: | View Download Statistic |
Actions (login required)
View Item |