Happy that your photos are safe online hidden behind aggressive watermarking? Maybe it’s time to reconsider. New Google research shows that a lot of watermarks, including those used by major stock websites, can be easily removed automatically by computers. But there’s a way to prevent it.
Using clever AI algorithms, it’s possible for a computer to zero in on the exact watermark and remove it from a photo as if it was rubbing away a smudge. Here are a couple examples Google shared using stock photos:
The news was disclosed in a paper, titled “On The Effectiveness Of Visible Watermarks,” that was presented at the 2017 Computer Vision and Pattern Recognition Conference.
Google has brought this to the attention of the photography industry in the same way that it would reveal information about software vulnerabilities to malware. There are some solutions, though, to make your watermarks “more robust to such manipulations.”
The removals work by identifying repeating patterns (such as watermarks) in a large collection of photos with the exact same watermark — you may have the same thing on your photos if you use an action to apply your watermark. The computer can then establish a rough estimate of the watermark and what exactly it looks like by viewing the image as noise, and the watermark as the target.
The original photo is then recovered by solving what Google calls a “multi-image optimization problem.” This involves separating the watermark (the foreground) from the photo itself (the background).
The optimization can produce “very accurate estimations” of the watermark’s own components and is able to deal with most watermarks seen on all kinds of photos. Here are more examples:
The solution is relatively simple, though. The problem lies in the fact that there is “consistency in watermarks across image collections.” So, to counteract the ease of removal, photographers need to somehow introduce inconsistencies for their watermarks. Even a subtle warp of your watermark in each photo is enough.
“We found that introducing random geometric perturbations to the watermark — warping it when embedding it in each image — improves its robustness,” said Google. Here are some examples showing how warped watermarks can’t be cleanly removed by the same AI system:
Here’s a 3-minute video explaining this research:
Thanks to this research, it’s possible that stock libraries will start introducing such random changes. Right now introducing geometric changes to your own photos isn’t easy to automate, but who knows – maybe Adobe will introduce such a feature to the Lightroom export window in the near future.