TY - GEN
T1 - Scene Text Transfer for Cross-Language
AU - Zhang, Lingjun
AU - Chen, Xinyuan
AU - Xie, Yangchen
AU - Lu, Yue
N1 - Publisher Copyright:
© 2021, Springer Nature Switzerland AG.
PY - 2021
Y1 - 2021
N2 - Scene text transfer for cross-language aims to erase the original scene text and generate another language text image into the original scene text image with the same style, including the style of fonts, colors, size, and background texture. Scene text transfer for cross-language is a challenging problem as the complicated background scene and a huge difference between languages, which demanding high-quality performance for both text transfer and text erasing. In this work, we propose a scene text transfer framework for cross-language which consists of three steps: regional text extraction, style transfer, and scene text combination. The regional text extraction is designed to crop the text region of a natural scene image and transform it to be a rectangle text image. In the second step, a style transfer network is proposed to retain the style of text image and transfer the text content. In the step of the scene text combination, our model combines the rendered text image with the original scene image to produce the final result. In the optimization part, we introduce a novel background consistent loss to improve the performance of background generation. Experiments demonstrate that our framework generates scene text images of higher quality than previous methods.
AB - Scene text transfer for cross-language aims to erase the original scene text and generate another language text image into the original scene text image with the same style, including the style of fonts, colors, size, and background texture. Scene text transfer for cross-language is a challenging problem as the complicated background scene and a huge difference between languages, which demanding high-quality performance for both text transfer and text erasing. In this work, we propose a scene text transfer framework for cross-language which consists of three steps: regional text extraction, style transfer, and scene text combination. The regional text extraction is designed to crop the text region of a natural scene image and transform it to be a rectangle text image. In the second step, a style transfer network is proposed to retain the style of text image and transfer the text content. In the step of the scene text combination, our model combines the rendered text image with the original scene image to produce the final result. In the optimization part, we introduce a novel background consistent loss to improve the performance of background generation. Experiments demonstrate that our framework generates scene text images of higher quality than previous methods.
KW - Dataset synthesis
KW - Generative Adversarial Networks
KW - Text style transfer
UR - https://www.scopus.com/pages/publications/85116888592
U2 - 10.1007/978-3-030-87355-4_46
DO - 10.1007/978-3-030-87355-4_46
M3 - 会议稿件
AN - SCOPUS:85116888592
SN - 9783030873547
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 552
EP - 564
BT - Image and Graphics - 11th International Conference, ICIG 2021, Proceedings
A2 - Peng, Yuxin
A2 - Hu, Shi-Min
A2 - Gabbouj, Moncef
A2 - Zhou, Kun
A2 - Elad, Michael
A2 - Xu, Kun
PB - Springer Science and Business Media Deutschland GmbH
T2 - 11th International Conference on Image and Graphics, ICIG 2021
Y2 - 6 August 2021 through 8 August 2021
ER -