Mask-Guided Font Style Transfer in Natural Scene
Jing Yang, Siyu Xia · 2022 41st Chinese Control Conference (CCC) · 2022
Font style transfer in natural scene is a challenging task, because it is necessary to maintain the content, color char-acteristics and background information of the font in the original scene. In this paper, we propose a mask-based method to change the font style in natural scene pictures. We design a two-stage method to complete this task. Firstly, the font part in the image is extracted through the image segmentation network for font and generate corresponding mask, at the same time, the feature information of the surrounding part of the font is reserved for the later font background fusion part. Next, the font style is converted through the generated mask to realize the transformation from one font to another. We have proved the effectiveness of our method through a series of experiments, and our method can be applied to tasks such as augmented reality.