4.5 Article

Scene text removal via cascaded text stroke detection and erasing

期刊

COMPUTATIONAL VISUAL MEDIA
卷 8, 期 2, 页码 273-287

出版社

SPRINGERNATURE
DOI: 10.1007/s41095-021-0242-8

关键词

scene text removal; text stroke detection; generative adversarial networks; cascaded network design; real-world dataset

资金

  1. National Natural Science Foundation of China [62102418, 62172415]
  2. National Key R&D Program of China [2019YFB2204104]
  3. Open Research Fund Program of State key Laboratory of Hydroscience and Engineering, Tsinghua University [sklhse-2020D-07]

向作者/读者索取更多资源

This paper proposes a novel end-to-end framework based on accurate text stroke detection for the scene text removal task. Experimental results demonstrate that the proposed method outperforms the state-of-the-art for locating and erasing scene text. Additionally, a new large-scale real-world dataset has been constructed to facilitate research.
Recent learning-based approaches show promising performance improvement for the scene text removal task but usually leave several remnants of text and provide visually unpleasant results. In this work, a novel end-to-end framework is proposed based on accurate text stroke detection. Specifically, the text removal problem is decoupled into text stroke detection and stroke removal; we design separate networks to solve these two subproblems, the latter being a generative network. These two networks are combined as a processing unit, which is cascaded to obtain our final model for text removal. Experimental results demonstrate that the proposed method substantially outperforms the state-of-the-art for locating and erasing scene text. A new large-scale real-world dataset with 12,120 images has been constructed and is being made available to facilitate research, as current publicly available datasets are mainly synthetic so cannot properly measure the performance of different methods.

作者

我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。

评论

主要评分

4.5
评分不足

次要评分

新颖性
-
重要性
-
科学严谨性
-
评价这篇论文

推荐

暂无数据
暂无数据