Paper Title:
Picking watermarks from noise (PWFN): an improved robust watermarking model against intensive distortions
Published on:
8 May 2024
Primary Category:
Multimedia
Paper Authors:
Sijing Xie,
Chengxin Zhao,
Nan Sun,
Wei Li,
Hefei Ling
Introduces denoising module to recover lost watermarks after noise layer
Adds SE module to encoder to fuse watermark pixel and channel-wise
Achieves state-of-the-art robustness under high noise levels
Matches performance of existing methods under low noise
Ablations validate superiority of proposed modules
Recovering lost watermarks using image denoising
This paper proposes a robust image watermarking model that introduces a denoising module between the noise layer and decoder in the typical encoder-decoder architecture. The denoising module reduces noise and recovers watermark information lost during attacks, improving robustness. Additionally, a SE module is added to the encoder to fuse watermarking information pixel and channel-wise, enhancing efficiency. Experiments show the model matches or exceeds state-of-the-art methods under high noise levels. Ablations demonstrate the value of each proposed component.
Embedding Watermarks in Stable Diffusion Models
Transformer for Joint Image Decoding and Denoising
Illuminating the Darkness: A New Vision for Low-Light Image Processing
Improving language model robustness through self-denoising
Associative Watermarking for Image Authentication
Detecting and evaluating watermarks in AI text generation
No comments yet, be the first to start the conversation...
Sign up to comment on this paper