Stock photos have always had a problem of people removing watermarks to effectively steal the images. This was made even more problematic after Google published a paper detailing how to automate the process. Fortunately, Shutterstock has managed to reverse engineer the algorithm and create a new removal resistant watermark.
Google’s algorithm relied on having access to samples of the watermark. The machines examine what it looks like and then is able to remove the watermark without doing too much damage to the underlying image. Of course, this relies on the practice of contemporary watermarks which are all the same.
Shutterstock’s solution is to create a watermark randomiser. Effectively creating a different pattern for the watermark every time it’s applied. The subtle variations prevent the algorithm from learning what the watermarks look like. Shutterstock is also including the contributor’s names in the watermark to make things even more difficult for automated systems.
This may not stop all watermark removal, but it does shut down one avenue for people who “borrow” stock images without paying for them.
[Source: The Next Web]