Photoshop’s Content-Aware Fill versus Watermarks

Written by *Enter New Author
|
Published on May 18, 2010
*Enter New Author
Adorama ALC

Is Content-Aware Fill a surefire watermark killer?

Content-Aware Fill is one of the hottest new features in Adobe Photoshop CS5. For example, a 5-minute sneak-peek demo on YouTube has been viewed over 3 million views in the past two months. (See all the sneak peek videos here!)

It is also one of the new features that has been generating a lot of discussion and concern, particularly in editorial circles, as I explored last week.

Additionally, many pro photographers have expressed concern that Content-Aware Fill is potentially a magical watermark killer: that the abilities that C-A-F may offer to the unscrupulous user in terms of watermark eradication are a serious threat and may lead otherwise honest individuals down a path towards willful image manipulation and unauthorized use.

However, I maintain my stance that there must be a conscious decision to choose to behave in an untoward manner, and that the ease with which new tools facilitate a dishonest act are mostly irrelevant. The intent must exist first for the tools to matter. To make a writing analogy once again: plagiarism is plagiarism, regardless of whether the pilfered words were manually typed in to a document and claimed as one’s own, or simply copied and pasted from the original source and claimed as one’s own.

With all this in mind, I set out to test Content-Aware Fill against a series of images from my collection of images on Photoshelter. My watermark on my Photoshelter images is pretty complex: It is a combination of vector and raster elements with varying grayscale values that overlays roughly one eighth of the image.

I chose a sampling of images from my collection of different visual complexities, ranging from repetitive patterns with deep depth of field to isolated subjects with crisp background separation.

I made a loose feathered selection around my watermark and clicked delete to bring up the Fill dialogue box and selected Content-Aware Fill at 100% blending as the option in every case. Let’s see how C-A-F did with these six images.

Swimming Minnows

All in all, Content-Aware Fill did do a very good job in eliminating the watermark in this shot. Of course, this image is nothing but a fluidic repetitive “pattern”, and it would likewise be very, very easy to fudge away the watermark with classic copy/paste or clone stamp operations.

Flower Boxes

Here we again have a very repetitive image, although there are competing patterns making this a more complex image. Making a lasso selection and clicking Content-Aware Fill was quick and easy, but C-A-F doesn’t nail the radial nor the grid pattern accurately enough to be considered a success. Manually copying, pasting, rotating and blending in pixels from other squares would kill this watermark a lot more believably.

Aerial Cemetery

Content-Aware Fill gets use close, quickly. But it will still require some manual touch-up to completely present a seamless altered reality sans watermark of this aerial shot of a cemetery.

Silhouette Spider

Here we have a complex contour lines and interesting positive and negative space in a two-color image. Content-Aware Fill more or less mangles it. It’d be much easier to use a combination of local selections, levels adjustments, and cloning/healing tools to lift the watermark here.

Seagull

Here we’ve got a sharp centered subject amid a repetitive background. Content-Aware Fill decided to repeat the pattern and in the process sticks the gull’s head where the wings should be.

Whale Tail

Again, we’ve got a main subject against a repetitive background. Again, the results aren’t perfectly seamless nor believable.

As you can see from these examples, it seems Content-Aware Fill does best when there is a lot of small pattern-type information in both the fill and sample areas of the image. We had varying degrees of failure and success attacking these watermarked images with Content-Aware Fill.

In other words, Content-Aware Fill isn’t an omnipotent watermark killer. It is just another tool in the Photoshop arsenal.

There are plenty of legitimate applications for Content-Aware Fill, just as there are plenty of legitimate applications for so many of the other copy/paste/heal tools and workpaths in Photoshop.

From a mathematical perspective, the pixels in a watermark are no different than the pixels in many other semi-transparent presentations within a captured frame: partial shadows on a subject’s face, light projected onto an uneven surface, partial reflections and/or glare through glass panes or on water surfaces, branded etchmarks on a wine glass that must be removed for generic stock sales, and so on. And for these usages and situations, it can be a good tool.

But again, it is just a tool.

And the decision to employ any tool for a dishonorable purpose must be consciously made on the part of the user.

To illustrate it another way, I can type from a printed source at about 30 words per minute, and I can copy and paste even quicker than that, but it would require a deceitful intent to claim these words of advice as my own:

If you must write prose and poems
the words you use should be your own
don’t plagiarise or take “on loans”
there’s always someone, somewhere
with a big nose, who knows
and who trips you up and laughs
when you fall.

Many will recognize these words as coming from the song “Cemetary Gates” by The Smiths, and it would be absolutely outlandish to try to claim otherwise.

But in reality all it took was a simple copy/paste from a lyrics site to put them in this article. The decision to correctly attribute these words comes from my personal sense of right and wrong.

Tools are just tools.

What are your thoughts on Content-Aware Fill? Let us know.