Researchers propose 'Signpost Watermarking' for layered image/video authentication
A new method lets multiple invisible watermarks coexist in images and videos without interference...
A new method called Signpost Watermarking trains imperceptible visual watermarks to coexist with other such watermarks. Previous work found that independently trained image watermarking models can work together, but that coexistence is more accidental than designed, leaving interference uncontrolled and potentially hurting decoding robustness or visual quality. This method explicitly optimizes for coexistence, first showing the same effect extends to video watermarking, then using a decoder-aware objective to improve coexistence for both images and videos. The result points toward signpost watermarks that signal the presence of independently deployed provenance systems, supporting layered provenance signaling for content authenticity and rights.
- Signpost Watermarking is the first method to explicitly optimize for visual watermark coexistence in both images and videos
- Prior methods relied on serendipitous coexistence, leading to uncontrolled interference and reduced robustness
- Enables layered provenance signaling where multiple stakeholders can embed independent watermarks without degradation
Why It Matters
Solves a critical challenge in digital content authenticity by enabling conflict-free multi-watermark systems for rights management.