Watermarking is like "tagging" a document by modifying it to carry a hidden information (often few bytes). The document looks like original, there is no attempt to make it unreadable or to restrict its usage. This technology belongs to steganography.
Watermark is used to fight piracy by inserting private information in the media file, which can identify its owner, buyer or creator for example. Lets use an example.
You purchase an audio track on a online store on which you have a user account. The audio track you get (mp3 file for example) is not a generic one but a slightly modified one with your name, user-id or other hidden in it. This file remains a fully playable and valid one with no restriction, there is just some modifications in it you can't eared.
You have not the right to spread this file you bought. If this file is found, freely available on a peer-to-peer network for example, the online store can analyse it to retrieve hidden data and to identify you as the owner, so maybe the one who initiate this illegal use of the file.
The same can be done for other file types. Data is hidden depending on the file format to generate a file usable and not different from the original when it is rendered (from a human point of view).
Here are items which can be used to compare efficiency of watermarking systems:
- modifications should not be perceptible for a human, the less perceptible the more efficient it is,
- it should not restrict legal usages at all,
- hidden data should be difficult to delete and should remain readable if file have been modified, resized, cropped, converted in another file format. The more difficult the watermark is to remove, the more efficient the watermarking is.
Also, for large media files, the faster it is to watermark the better it is. It allows buyers not to wait hours their legal downloads.