Digital watermarking technology has been maturing since an explosion in interest began in the mid 1990's.  While electronic watermarking dates back to the 50's and a handful of ideas were tried in the late 80's and early 90's, it wasn't until the mid to late 90's that the technology started to make significant leaps.  For some applications, digital watermarking technology has matured enough to be an interesting tool for developers.

As with many blog entries, this one was inspired by a recent news item.  Civolution has acquired Thomson's watermarking business unit (Civolution Press Release).  This marks the end of Thomson's 4 year experiment in selling digital watermarking products and is yet another example of a large company entering and then exiting the digital watermarking marketplace.  There is much to talk about here: the history of Thomson's efforts, the genesis of Civolution as an exit strategy for Philips, the domination and then refocusing of Digimarc, examples of market acceptance and market rejection of digital watermarking technologies, etc.  I will hold my tongue on all of those interesting topics for the moment (coming back to them in a series of posts to follow) and first provide some background about the technology itself. 

A digital watermarking system has 2 components: a watermark embedder and a watermark detector (the embedder is sometimes implemented as 2 components).  We'll restrict the discussion to the watermarking of multimedia items (audio stream, video stream, or still image), but there is technology to watermark text, 3D models, and other digital objects as well.  The embedder operates on such a multimedia item, or work (terminology from copyright law, as in "a work of art").  The embedder also takes some watermark information as input.  This might be a serial number or an index into a database, or a time stamp - essentially any side information about the work.  As its name implies, the embedder embeds the watermark information into the work.  The embedding changes the actual pixels or sound samples of the work rather than putting the info in a header.  With a good digital watermarking technology, the watermarked version of the work is perceptually indistinguishable from the original version.  In other words, the added data is invisible and/or inaudible.

The second component of a watermarking system, the detector, operates on a watermarked work and extracts the embedded watermark data.  Together, these two components allow the work itself to act as a communications channel.   Applications of watermarking are limited only by a developer's imagination.  Some commercial applications include copyright protection (embed copyright information), broadcast monitoring (advertisements are watermarked with a unique ID and detectors "watch" the broadcast looking for and logging advertisements), A/V sync, invisible hyperlink (recovered with a camera or scanner), authentication (watermark contains a hash of the work that will change if the content is tampered), and forensic analysis (sensitive works are watermarked with the ID of the person receiving the copy and can thus be used to identify the source of a leaked copy).  Watermarks have also been used in the toy industry.  The VEIL watermarking system was used to embed watermarks into the video portion of a children's television program.  A light sensitive device placed near the television recovered the watermark data and passed instructions, via high-frequency IR, to a toy.  The result was that the toy could respond to instructions embedded in the television program.

Pretty cool technology with some pretty cool applications.  Historically, there has been a disproportionate focus on security applications, but I suspect the most interesting apps to come will not be security apps.  I suspect also that the readers of this blog will come up with these creative applications of digital watermarking.