Street Photography Magazine

Invisible Watermarks | Tool Test

Digital watermarks were a hot topic at the

turn of the 21stcentury, but have since

disappeared somewhat from the public eye.

This test only looks at tools that have been

updated within the last two years. All other

legacy tools are listed in the table on page 95.

The following sections concentrate on tools

that offer digital photographers genuine help

when protecting their intellectual property.

Only Digimarc and SignMyImage really fulfill

our test criteria, and both offer their own

crawler service.

The online Photopatrol service was an

attempt to offer a cloud-based photo tagging

service based on checksum methodology. The

service hasn’t been updated since it was

acquired by the DAPD news agency, which

now uses it as an internal tool as well as selling

its services to other agencies. The Photopatrol

crawler searches for metadata and can thus

only find images that include appropriate

keywords. Because the service is not currently

supported, we left it out of our test.

SignMyImage

SignMyImage (SMI) is available as a

standalone program, as a Photoshop

plug-in or as a command-line utility. The

manufacturer’s own crawler service is called

Image Spider and has to be activated by

purchasing an annual subscription in addition

to the program itself. There is also a free trial

version of the program available that allows

you to test its functionality but with the

restriction that it produces visible watermarks

in the resulting images. Manufacturer ADP

tools (www.adptools.com) also offers an

end-to-end photo sales and protection suite

called Vericuff.

We conducted our test using the 4.0

standalone version of the software but,

because this version appeared somewhat

unfinished, we also tested the older, more

stable 3.52 version. The results in the table on

page 92 relate exclusively to the 4.0 version.

Compatibility-wise, the newer version no

longer supports 16-bit RGB TIFF and 8-bit

grayscale images, whereas the older version

does. The user interface has also lost some

ground in the new version, and many

functions that were easily accessible in the

previous version are now hidden away in the

tool’s menus. Both versions are quite slow and

took as long as 45 seconds to embed a

signature in a high-resolution sample image

using our test system (3.3 GHz Intel Core i3

2120 with an SSD and 16 GB of RAM) with all

cores active. Digimarc took about 15 seconds

to perform the same task, which is better, but

still not particularly fast.

Rather than spreading watermarks around

the entire image area, SMI embeds them in

selected areas and tells the user where these

are once it has completed the process. These

areas are recognizable by their slightly convex

appearance in the difference image in

Photoshop, and can cause a visible reduction

in image quality in monotone areas. SMI only

partially protects images against serious

cropping, and if you crop outside the marked

areas, the watermark will be destroyed. SMI

signatures don’t survive conversion to

grayscale and are also susceptible to damage

by a variety of image processing tools,

including standard processes such as noise

filtering and distortion correction. Reducing

image size to 25 percent or less also

deactivates the signature, as does rotating an

image through just 0.3 degrees using

Photoshop.

At the end of the day, SMI offers great

value for money but still embodies all the

familiar pros and cons of checksum

signatures. The software, bundled with a

year’s subscription to the crawler service,

costs just US$32 and, unlike Digimarc, you

Tools for Creating and Managing

Invisible Watermarks

A few years ago there was a wide range of digital watermarking tools available.

Time and technological advances have taken their toll, and some of the high-flyers from

yesteryear have now disappeared from the market. There are currently just two services

offering functionality that is robust enough for use in today’s online world.

Digital watermarks are hidden amongst

the pixels that make up an image and are

stored in the form of artificial luminance

or color noise that slightly alters the

original image data. A high-quality

watermark is not visible to the naked eye

but is robust enough to survive deliberate

attacks. Altering image data is, in itself, a

trivial task that can be performed using a

wide range of free and commercial image

editing programs. The problem with

removing watermarks using basic editing

techniques such as noise filtering, scaling

or distortion is that this also destroys

much of the image data, thus rendering

the resulting file useless. We tested the

tools detailed here by embedding

watermarks at their default strength.

Stronger watermarks are more robust but

have more obvious effects on the images

they are designed to protect.

We used everyday editing techniques to

subject our test images to real-world tests

but without doing them too much

damage. For example, we converted RGB

images to grayscale or a lower bit depth.

Watermarks based on changes in

luminance data generally survive these

tests quite well, whereas the nature of the

process makes those based on additional

color noise data much more susceptible to

damage and/or deactivation.

A watermark that is worth its salt should

survive all common image editing steps,

including distortion correction, noise

filtering, unsharp masking and rotation.

When we tested digital watermark

software three years ago, many of our

test candidates were particularly weak

when it came to applying rotations, so

we were interested to see if the

manufacturers had done anything to

improve the situation since then. Image

size reductions are also quite common –

for example, when you scale a

high-resolution image for online

publication as a thumbnail. We used

Photoshopfor most of our tests but, due

to Adobe’s strong cooperation with

Digimarc, we also performed some

additional spot checks usingGIMPand

Nikon’s Capture NX 2.

Our Test Methods