The Digital Watermark

(This column first appeared in the February 20, 1996 issue of PC Graphics Report)

Background Info

Last week, scientists at NEC Research Institute (Princeton, NJ) announced they have developed a digital watermarking method for use with audio, image, video and multimedia data, which they claim is highly effective in protecting copyrights of images and music on the Internet.

According to NEC, conventional cryptographic systems permit only valid keyholders access to encrypted data, but once such data is decrypted there is no way to track its reproduction, and therefore, it provides little protection against data piracy. A digital watermark is intended to complement cryptographic processes. It is an invisible identification code that is permanently embedded in the multimedia data.

Unlike previous methods, NEC’s method places a watermark in perceptually significant components of a signal, which makes the removal of the watermark virtually impossible. It is well known that modification of these components can lead to perceptual degradation of the signal. To avoid this, a watermark is inserted into the spectral components of the data using techniques analogous to spread spectrum communications. While not visible or audible to human senses, the watermark is robust to common signal and geometric distortions such as digital-to-analog and analog-to-digital conversion, resampling, and requantization, including printing and compression, and rotation, translation, cropping and scaling. The same digital watermarking algorithm can be applied to all three media (audio, image and video) with only minor modifications, making it especially appropriate for multimedia products. Retrieval of the watermark unambiguously identifies the owner, and the watermark can be constructed to make counterfeiting almost impossible. Attempts to remove the watermark from an image will result in a noticeable degradation in image quality well before the watermark is lost, thereby rendering the image useless.

Because the watermark allows unique identification of copyright owners, buyers and distributors, NEC believes that it provides a strong deterrent to illegal copying. NEC scientists believe that this watermarking technique is fundamental enabling technology that will help create a secure business environment for copyright holders using the internet.

The Mark’s Where It’s At
I found this announcement significant for several reasons. First, I think that protecting intellectual property on the Internet is a major problem. Second, I’m an amateur photographer who wants to promote his photographs (and my wife’s oil painting) via the Web, but I haven’t done so yet because I’m trying to find a real way to protect our copyright when I post images of our works.

I tracked down Ingemar Cox, a senior research scientist at NEC Research Institute, and discovered that this new watermark technology is still in its early stages. So far, it’s only been testing extensively on black and white images and audio, with testing on color images and video scheduled shortly. They expect no problems.

The watermark is composed of a bit pattern distributed throughout the data based on noise theory, and is calculated using discrete cosine transforms (I used to know what that meant). The watermark apparently causes no visual aural degradation of the image or sound stream, as it is stored in an imperceptible region of the data. The watermark is also close to indestructible, because it’s not purely digital in nature, but a digital representation of analog data. This means that a significant number of bits that make up the whole watermark can be altered without dramatically affecting the ability to uniquely identify the watermark.

Ingemar told me that with a gray scale image, encoded with the watermark, they printed it out dithered, photocopied it, faxed the copy, and found that the resulting fax still contained enough of the watermark to uniquely identify it. The watermark will even survive compression which is loss-based, such as JPEG, MPEG, and virtually all other efficient means of video and audio compression. Very cool!

My first impression, after reading the release from NEC, was that the watermark was a way of uniquely identifying the owner of the watermark. That’s not the case. Actually, a unique watermark is applied to the image (takes mere seconds, if that) and the watermark signature is then stored. Later, if an infringement of the copyright is suspected, the image in question can have its watermark extracted (even if it appears in printed form!) and compared to the original. Even better, over a billion totally unique watermarks can be encoded in an image – as Ingemar put it, probably one for every man, woman, and child in the world. This means that you could create a unique watermarked image for everyone who downloaded it off your Web site, on the fly, storing a copy of their watermark signature along with their name or other contact info. Then, if infringing use of the image or audio file is found, it can be tracked uniquely back to whom it was issued.

NEC doesn’t yet have a time table of when the watermark technology will be commercialized or licensable, but they are moving in that direction. I for one, can’t wait.

Since I wrote this column, I’ve had several people point me to other types of Digital Watermark technology on the Web. For your benefit, here are a couple of new links:

The Dice Company:
Digimarc Corporation:
HighWater Signum Ltd.:

Jake Richter
March 17, 1997