About that photo of George Zimmerman’s bloody and “broken” nose… part 3
Even though it seems nobody other than me thinks the Wagner photo is really a big deal, since I still do, I shall plunge ahead with the third in my series of posts about it.
Again, my belief that this photo has been edited is based most strongly, and I would think persuasively, upon the anatomical anomalies I perceive when comparing it to other verified photos of Zimmerman taken later the same evening. That’s all presented in the part 1 post.
CAVEAT: This post will be longer, much geekier, and less conclusive. Unless you have some interest in the workings of new media technology you may well find it unilluminating.
As I’ve already noted: though I’m a fairly experienced Photoshop user, I have no expertise in the specific skills of photo-retouching, nor in specific methods used to ferret out retouching by a close examination of the pixel patterns in the image raster. The anatomical issues to the side, if I magnify the area in the Wagner photo that I think has been diddled, I don’t see any obvious signs of manipulation.
But as someone on FLLB observed the other day (FL himself IIRC), “absence of evidence isn’t evidence of absence.” I know enough about Photoshop to know that most people who use it only ever master a fraction of it’s abilities, and I believe that given enough time, a truly gifted and expert photo editor equipped with the right software can accomplish just about anything. So I’m not going to make an argument that any of the detail seen above should or should not be taken as authentic on technical grounds. That kind of thing is over my head.
Instead, I want to address three technical characteristics that have, or at lest appear to have, been changed in the released version of the Wagner photo from the original that would have been recorded by Wagner’s iPhone 4s. That is, SOMBODY, did SOME THINGS to this picture. While I can’t directly connect those things to altering the shape of Zimmerman’s nose, I will offer hypotheses on how they may suggest a photo editor was at work.
The three issues are: 1. Resolution, 2. Aspect Ratio, and 3. Color Saturation.
A digital photograph is made up of a grid of dots, called a ‘raster.’ Each dot (called ‘pixels’), each square in the grid has a color value defined by three variables: brightness, hue (position on the color spectrum) and saturation (richness or intensity of the hue element in comparison to neutral). The more boxes in the grid recording any given image, the more detailed the photo will be. The dimensions of the raster are referred to as the photos ‘resolution.’ All other things being equal (which they aren’t but that’s another story…) the bigger the better. A photo editing program like Photoshop can resize a digital photo, including blowing it up to create a higher resolution, but it can’t create detail that wasn’t there to begin with in the process. On the other hand, when you use Photoshop to make a photo smaller, decreasing the resolution, you inevitably lose detail tat can’t be recovered by resizing the reduction back up again. The original detail level exists only in the original digital file.
Wagner took his photo with an iPhone 4s. Now, unlike many digital cameras, which can be set to take photos in different resolutions, the iPhone 4s only takes photos in one resolution: 1936 x 2592 pixels. That ‘5 Megapixels” in digital camera advertising jargon, pretty good resolution for a cellphone camera. However, the Wagner photo as released is only 650 x 870 pixels. In other words, it has been reduced to about 1/9th of it’s original area.
That reduction, in and of itself, is hardly nefarious. Digital photos are typically reduced down around that size for posting on the Web. The lower resolution makes for a much smaller file, taking up less bandwidth, loading faster and so on. But it would also be much easier to hide traces of manipulation in a low resolution, coarse detail image than a high resolution, fine detail image. And we don’t know who/when/how of the Wagner photo being reduced.
To give you an idea of the difference in resolution, here is the Wagner photo as released on the web, and an unreduced photo of a theatrical mock-up created to resemble the Wagner pic, photographed with an iPhone 4s (just like the one Wagner used). Each photo should expand to full size when you double-click on it.
Note how much more detail exists in an original iPhone image as compared to the Wagner image.
Now, if Wagner sent a full resolution 1936 x 2592 image file to the investigators, and some FDLE media tech just reduced it to put it on the Web, then the odds of manipulation would go down. It would still be possible, but it would be extremely difficult and time consuming to fake a nose structure that would pass muster at 5 Megapixel level detail, especially given that this is original resolution.
This is because the act of reducing isn’t just a matter of losing detail, it also involves changing the way detail is represented by the pixel pattern in the raster. Let’s say you started with a 650 x 870 image and reduced it to 600 x 803. The software has to figure out what values to assign each pixel in the new raster, Since the dimensions don’t divide into each other evenly, the program reads the values of all the pixels in the old raster that touch the area of the photo where the new pixel will be, and blends the values in a complicated formula that accounts for the relative distance and alignment of the old pixels with the new. But blend it does. How this blending occurs can be adjusted via several variables inside the software, and slightly different results can be obtained by doing the reduction in a series of stages as opposed to all-at-once.
In short, not only would it be easier to fake something on a low rez version than on a high rez version of a given photo, but manipulations made at a higher resolution can be effectively made less apparent by the act of reduction itself. Both of these techniques could be combined as well: editing at high rez, blending by reduction, followed by pixel editing at low rez to remove any telltale artifacts.
Again, I’m not a skilled photo retoucher. There are definitely people who can do this kind of thing way better than I can. But look what even I was able to do in about half-an-hour: I moved mock-Zim’s right eye (screen left) about 3/8″ closer to his nose using the original iPhone file. I did some blending with the brush-eraser set at medium opacity, and some removal of redundant lines with the rubber-stamp clone tool, also at medium opacity. Then I reduced it a couple steps, added a couple small tweaks with the clone and smudge tools, and reduced it a few more steps until reaching 650 x 870, same as the Wagner photo.
The editing is rather obvious if you look at the whole picture since the face is now way out of symmetry. I can tell where the traces of my editing are because I know what I did, but can you? Ignoring the anatomical anomaly, could you swear this was faked on the basis of raster realism alone? If you are that clever, remember a newb did this in 30 minutes. Imagine what a pro could achieve with almost three weeks to work on it.
2. Aspect Ratio
The aspect ratio of a digital image is the fraction formed by dividing the length of it’s long side by the width of it’s short side. So an 870 x 650 photo has an aspect ratio of 1.34. This also happens to be the aspect ratio of 1936 x 2592. Normal reduction or enlargement of an image doesn’t change the aspect ratio. You can change the aspect ratio of image by cropping just one side, or cropping different sides disproportionally. But you can also change the aspect ratio WITHOUT cropping, by scaling the image disproportionally along the horizontal and vertical axes. In other words, digital photos can be squeezed or stretched. Here is the Wagner photo squeezed sideways 10% without altering the height.
Such an image is referred to as ‘anamorphic,’ which is just tech talk for ‘the proportions are not what they were in the original.’ If an image is just a little bit anamorphic, it can be hard to tell at first glance. You might not see anything wrong with the 10% side-shrink version of the Wagner pic above if you didn’t have something to compare it to. The only reliable way to tell if an image has been made anamorphic is to compare the shape and proportions of something in the image to it’s known real-world shape and proportions.The usual go-to test in checking for proper aspect-ratio is to look for something in the image you know to be perfectly round — a side view of a wheel for example — and check to see if it’s actually evenly round, or has been elongated into an oval along either the horizontal or vertical axes.
Those of you with sharp eyes who have compared the released version of the Wagner photo with GZ’s SPD photos may have already observed that GZ’s head appears somewhat skinnier in the former than it does in the latter. Some of this may be the high camera angle accompanied by the close perspective coming from the wide-angle lens incorporated into the iPhone. But I think the Wagner photo as released IS slightly anamorphic, squeezed sideways or pulled out vertically by about 3%.
I can’t be definitive about this the image doesn’t contain a good enough known shape for comparison. Because of the different camera angles, we can’t compare GZ’s head in this picture to his head in the SPD pictures. The closest things in the Wagner photo that should conform to a known shape are the irises of his eyes, which should, of course, be perfectly round. In the image below I have placed a round black circle over a blow-up of GZ’s left eye from the Wagner photo. Unfortunately, since the iris is partially covered by the eyelid, I can’t say that I’ve gotten the circle centered exactly. But I’ve done the best I can, and if you look closely, you’ll see there’s more of the iris showing below the circle than on the sides. That is, the iris is slightly oval, slightly taller than it is wide.
Here is a side by side comparison of the Wagner photo as distributed on the left, and the same image but with the horizontal axis expanded by 3% on the right. To my eye, the image on the right looks more natural, and the one on the left appears slightly squeezed. (as always, double click to enlarge…)
Ok, so I could be wrong about the Wagner pic being a little anamorphic, but you may be asking, well, what difference does it make if I’m right anyway? The answer is that if indeed the image proportions have been altered that is a sign of intentional manipulation. The alteration could not have happened by accident. Here’s the thing… It would be fairly odd for a computer klutz resizing the photo to mess up the aspect ratio of the smaller file, making it anamorphic, but it’s not inconceivable. But then the aspect ratio would be different from the original. And the aspect ratio of the Wagner photo is exactly the same as the larger iPhone original. So the only way for the image to be anamorphic AND in the same aspect ratio as the original would be for it to have been both cropped AND squeezed-or-stretched, and done so with precision so the math would work out.
Which brings us to the next obvious question: why would anyone go to the trouble of performing this particular manipulation? Since the change is small, any benefits are also small. But the change-proportions-and-crop (or vice versa) routine is easily done and would only take about a minute, so any gain, however small would be worth the time. I can think of two reasons to perform this alteration. First, making George’s head skinnier makes him seem weaker and more vulnerable. Second, stretching or shrinking one dimension of the photo would be another operation — like the resizing discussed above — that causes Photoshop to rewrite all the pixels in the raster, again with the potential to hide any artifacts from editing/compositing work, especially if performed multiple times.
This one is pretty simple. The color in the Wagner photo is awful. GZ’s skin has a yellow cast, his gray shirt appears light blue, and all the colors are pumped up well beyong their natural intensity. Some of this is just part and parcel of the iPhone camera, which has pretty lousy color rendition (when compared to a ‘real’ digital camera anyway). Not only are the hues typically less than accurate, but both the brightness contrast and the color contrast are poor. Still, while the iPhone camera may be bad, it’s not THAT bad. The mock-Zim photo was taken with the same kind of equipment and under the same conditions as the Wagner photo: iPhone 4s / flash / subject seated in open car door with interior light on / some ambient street light… But the color in the mock-Zim photo looks a lot more natural than the Wagner pic. Here’s a split screen:
Again, tweaking up the colors in Photoshop takes only a few seconds, and is relatively easy to do. The increased saturation makes the photo seem more dramatic, with the blood coming from the nose becoming especially vivid. (Mock-Zim’s ‘blood’ was drawn on with a red Sharpie, and doesn’t quite match the tone of real blood… the model was not up for actually bleeding for this little experiment.)
Adding it up
Unless an M.D. with expertise in facial injuries weighs in supporting my take on the anatomical anomalies posed by comparing the Wagner and SPD photos, I can’t say I’ve established beyond reasonable doubt — even for myself — that the Wagner image has been ‘Photoshopped.’ But I think the evidence is far beyond trivial. Except for the apparent massive deviation of cheek and nose bones, you could fairly say that everything little thing I’ve noted about the Wagner photo is just that — a little thing. But these little things are all going the same direction, and they combine with a story of the photo’s history from Michael Wagner that I find just too convenient. He takes the only photo of the injuries of a man who has just shot an unarmed teenager to death, takes it home, makes the effort to upload it to his computer, and then forgets he has it? Really? And why does he erase the original of a key piece of evidence in what is obviously a major case from his phone. The shooting draws immediate attention from Florida media, and soon the attention of the national media, and yet it’s three weeks before he remembers he took the only photo of Zimmerman’s bloody nose? … I suppose it’s possible, but if this doesn’t arouse your skepticism I’ve got a friend in Nigeria who could use your help with a little matter of some frozen funds…
The Bigger Picture
This goes beyond the question of George Zimmerman’s guilt or innocence. It goes beyond the question of whether law enforcement officers conspired — perhaps with the assent or even at the direction of their superiors — to protect a ‘friend of the force’ who was facing serious criminal charges. It goes beyond the question of whether State prosecutors allowed tainted evidence to be entered at trial without challenge in order to avoid implicating a police department in a scandal.
It goes to the larger issue of how digital media are used to establish ‘fact’ within the legal system, and anywhere else in our society where the stakes are high. The progress of the Wagner photo into court-accepted evidence is simply unacceptable. It doesn’t matter whether it was actually manipulated. It could have been, and that is enough for deep concern, if not outright outrage. No forms of digital media should be granted evidenciary status unless they are presented in the original format in which they were recorded, and with a clear ‘chain of custody’, all properly documented. This scrutiny should not be left to adversaries on one side or the other to invoke by challenge. It should be demanded as a prerequisite by law.
And this is not just a matter of vigilance toward possibilities of intentional manipulation. All audio/visual technologies distort reality to some degree — hiding things that are actually ‘there,’ and adding things that actually aren’t. Much of the time these distortions turn out to be trivial in terms of the practical purpose for which the media are intended. But when the stakes are high, little distortions can mean a lot. Media technologies of all sorts have proliferated wildly. Vast numbers of folks walk around carrying devices that can record sound, capture still images, and even capture moving images. Security cameras seem to be everywhere as well. The digital a/v documents in the Martin/Zimmerman case are absolutely typical of the contemporary technological presence in everyday life: cellphone audio compressed into .mp3s; cellphone camera images reduced into low rez .jpegs; lo-fi surveillance video rephotographed off a computer screen. If all technologies carry possibilities of distortion, these sub-optimal technologies (to put it nicely) are a veritable minefield of potentially misleading perception. We all must become better informed about what the output of these devices CAN tell us, and what it CAN’T, and why. But such knowledge and discretion is crucial for decision makers in policy, law enforcement, and mass communication. And sadly, this wisdom seems almost wholly lacking in those domains.