-
[pdf]
[bibtex]@InProceedings{Wang_2022_ACCV, author = {Wang, Baowei and Wu, Yufeng}, title = {Staged Adaptive Blind Watermarking Scheme}, booktitle = {Proceedings of the Asian Conference on Computer Vision (ACCV)}, month = {December}, year = {2022}, pages = {1812-1827} }
Staged Adaptive Blind Watermarking Scheme
Abstract
In traditional digital image watermarking methods, the stren-
gth factor is calculated from the content of the carrier image, which can
find a balance between the robustness and imperceptibility of encoded
images. However, traditional methods do not consider the feature of the
message and it is also unrealistic to calculate the strength factor of each
image separately when faced with a huge number of images. In recent
years, digital image watermarking methods based on deep learning have
also introduced the strength factor. They assign the strength factor of
each image to a fixed value to better adjust the robustness and imper-
ceptibility of the image. We hope that the network can choose the most
appropriate strength factor for each image to achieve a better balance.
Therefore, we propose a staged adaptive blind watermarking scheme. We
designed a new component - the adaptor, and used two stages of training
by training different components in different stages, and improved the
robustness and imperceptibility of watermarked images. By comparing
the experimental results, our algorithmic scheme shows better results
compared to current advanced algorithms.
Related Material