Google research questions the effectiveness of visual watermarks
Aug 24, 2017 by CGPress Staff
0
|
Tali Dekel and Michael Rubinstein have published a research paper questioning the effectiveness of visual watermarking to protect images. In particular watermarks that are unchanged and used on multiple images such as those seen in stock libraries are vulnerable to detection and removal using automated algorithms. To counter this risk the research suggests that watermarks should introduce subtle randomised changes per image. Simply moving the watermark or varying the opacity was not enough to fool the algorithm, but random geometric perturbations did improve their robustness. You can read more on Google’s research blog.