An independent expert in the field of image forensics, Hany Farid, has analyzed the original Raw file, compared it to the prize-winning JPEG file, and concluded that “all of [the pixels] are exactly in the same place.” He also says that the final photo has experienced “a fair amount of post-production” (as in, dodging, burning, etc.), which probably explains a lot of the seemingly incredible lighting in the image.
It turns out that the 2013 World Press Photo of the Year — the largest and most prestigious press photography award — was, in actual fact, a fake. The World Press Photo association hasn’t yet stripped the photographer, Paul Hansen, of the title, but presumably it’s just a matter of time. Rather than discussing the politics of photo manipulation, though — is it faked, or is it merely enhanced? — we’re going to look at how Hansen seemingly managed to trick a panel of experienced judges with his shooping skillz, and how a seasoned computer scientist spotted the fraudulent forgery from a mile off.
The photo, dubbed Gaza Burial, was purportedly captured on November 20, 2012 by Paul Hansen. Hansen was in Gaza City when Israeli forces retaliated in response to Palestinian rocket fire. The photo shows two of the casualties of the Israeli attack, carried to their funeral by their uncles. Now, the event itself isn’t a fake — there are lots of other photos online that show the children being carried through the streets of Gaza — but the photo itself is almost certainly a composite of three different photos, with various regions spliced together from each of the images, and then further manipulation to illuminate the mourners’ faces.
This revelation comes from Neal Krawetz, a forensic image analyst. There were two main stages to the analysis: First an interrogation of the JPEG’s XMP block, which details the file’s Photoshop save history, and then pixel-level error level analysis (ELA). To begin with, the XMP data shows that the original, base image was converted from Raw format and opened in Photoshop on November 20, 2012 (the same date that it was taken). Then, on January 4, 2013, the XMP block shows that a second Raw image was opened and added to the original. An hour later, a third image was spliced in. Finally, 30 minutes later the photo chimera was actually saved to disk. The January 4 date is interesting because it shows that the final photo was only edited a couple of weeks before the January 17 submission deadline, not soon after original photo was taken in Gaza — in other words, it was probably edited specifically for the contest.
2013 World Press Photo of the Year: Gaza Burial, by Paul Hansen, subjected to ELA analysis by Neal Krawetz
The next step is error level analysis. ELA basically compares the error level of pixels that have been modified by the JPEG compression algorithm (low amounts of change), and pixels that have been modified with photo manipulation (higher change). In the image above, which has been subjected to ELA, we see clear markers that are consistent with the photo’s spliced-and-manipulated history. Regions that have only been subjected to normal JPEG compression should have faint red/blue patches, while white patches show areas that have been subject to other forces. The bright white edges are caused by Photoshop’s sharpening algorithm — but the other bright white regions are likely due to extensive manipulation. Take a look at the man on the far left, carrying the child’s feet — his magically, digitally illuminated face is clearly shown on the ELA map. In fact, almost every face in the picture has been brightened, as have the children’s shrouds.
The final nail in the coffin is good ol’ shadow analysis. At the time the photo was taken — 10:40am, in the winter — the sun should be fairly low in the sky. The shadows on the left wall are consistent with a sun location (shown below) that should cast deep, dark shadows on the mourners’ right sides — but, as you can see, those magical light rays seem to be at work again.
Basically, as far as we can surmise, Hansen took a series of photos — and then later, realizing that his most dramatically situated photo was too dark and shadowy, decided to splice a bunch of images together and apply a liberal amount of dodging (brightening) to the shadowy regions. For what it’s worth, Hansen claims that the light in the alley was natural — and to be fair, sometimes magical lighting does occur. I think most of you will agree, though, that the photo simply feels fake — there’s just something about the lighting that sets off a warning alarm in your brain. As for why World Press Photo didn’t forensically analyze the photo using freely available, advanced, accurate analysis tools such as FourMatch or FotoForensics… who knows.
Oh, I forgot to mention the best bit: Hansen was meant to provide the Raw file for his winning photo, as proof that he didn’t significantly modify the final image — but so far, he hasn’t.
When is an image fake, and when is it merely enhanced?
The bigger discussion, of course, is whether Gaza Burial is actually fake — or just enhanced to bring out important details. This is a question that has plagued photography since its inception. Should a photo, especially a press photo, be purely objective? Most people think the answer is an obvious “yes,” but it’s not quite that simple. What if a photo is perfect, except that it’s taken at an odd angle — can you digitally rotate it? What about cropping? What if there’s dust on the lens/sensor/film — can you digitally remove it?
Perhaps most importantly, though, cameras simply don’t capture the same gamut of color or dynamic range as human eyes — a photo never looks the same as the original image perceived by your brain. Is it okay for a photographer to modify a picture so that it looks exactly how he remembers the scene?
