Watermarking

A digital watermark is a signal permanently embedded into a multimedia content (audio, image, video, text) that can be detected or extracted later by means of computing operations to make an assertion about the data.

To embed watermark information in the content, digital watermarking (or data hiding) techniques apply minor modifications to the host data in a perceptually invisible manner. The hidden information can be retrieved afterwards from the modified content by detecting the presence of these modifications. By means of watermarking, the information is hidden directly into the host content in such a way that it is inseparable from the data itself.

Such technology is useful in all the applications where it can be useful to associate some additional information (metadata) to a multimedia document: thanks to data hiding, this metadata can be inserted as a code directly into the content and not into its header. Examples of possible applications are: copyright protection, content labeling and hidden notations, authentication and integrity verification, secure and invisible communications.

The IAPP research team started to work on digital watermarking technology since 1995. The main experience on this area concern the design of watermarking algorithms belonging to the following classes:

  • Robust watermarking techniques for images and video sequences: they allow to hide some data useful for proving the content ownership and then to track the copyright violations; the embedded watermark is robust, that is it resist to a large set of possible manipulations applied to the content; our main scenario has been represented by Cultural Heritage applications.
  • Fragile or semi-fragile watermarking techniques for the authentication of JPEG or uncompressed images: these techniques allow to hide into an image some information useful to prove subsequently its authenticity. It is possible to assure that an image has not been tampered, and in some techniques also to locate the manipulations occurred that altered the original content of the image; our main scenario has been represented by banking applications.
  • Reversible watermarking techniques: these methods embed a watermark to authenticate images; when the watermark is detected, it is possible to restore the image to its original form by removing the watermark and replacing the original image data which had been overwritten; our main scenario has been represented by biomedical applications.

Research group:

Research area: