Visible Watermark Removal with Dynamic Kernel and Semantic-aware Propagation
Xing Zhao (Shanghai Jiao Tong University), Li Niu (Shanghai Jiao Tong University),* Liqing Zhang (Shanghai Jiao Tong University)
The 33rd British Machine Vision Conference
Abstract
Visible watermarks are designed to protect the copyright of images. Inversely, visible watermark removal studies how to enhance the removal resistance of visible watermarks. The existing watermark removal methods have achieved competitive results, but they cannot cope with diverse watermarks and complicated semantics very well. For the problem of diverse watermarks, we propose watermark-specific dynamic kernel, which can detect and remove watermark adaptively considering the specific properties of different watermarks. For the problem of complicated semantics, we develop semantic-aware propagation, which aims to reconstruct the corrupted pixels by borrowing information from semantically similar pixels in the neighboring region. We conduct ample experiments on two benchmark datasets, demonstrating that our method outperforms previous methods.Video
Citation
@inproceedings{Zhao_2022_BMVC,
author = {Xing Zhao and Li Niu and Liqing Zhang},
title = {Visible Watermark Removal with Dynamic Kernel and Semantic-aware Propagation},
booktitle = {33rd British Machine Vision Conference 2022, {BMVC} 2022, London, UK, November 21-24, 2022},
publisher = {{BMVA} Press},
year = {2022},
url = {https://bmvc2022.mpi-inf.mpg.de/0106.pdf}
}
Copyright © 2022
The British Machine Vision Conference is organised by