I didn't know what "Error Level Analysis" was. So:
'Error level analysis (ELA) works by intentionally resaving the image at a known error rate, such as 95%, and then
computing the difference between the images. If there is virtually no change, then the cell has reached its local
minima for error at that quality level. However, if there is a large amount of change, then the pixels are not at their local minima and are effectively “original”.'
Just consulted my biologist brother, who was worked extensively in the field in BC, and he doubts the photo is real:
- The wolf is huge, much larger than anything known to stalk the woods of BC (both of us are from BC).
- The amount of fur suggests that it's wildly out of its natural habitat - not one of the common wolf species, and probably something more Arctic than temperate.
- From his personal experience, wolf eyes reflect green, not yellow. Furthermore, you will only ever get the shine if you shine a light at it - like a flash - but the flash on the camera in question could not reach that far. In fact, the flash on that camera was not fired, which as a photographer I can corroborate. There is no way a flash was used in the making of this picture.
- There's some weird clipping in the wolf, and a fairly distinct lack of shadow, but neither of us are Photoshop experts, so that's neither here nor there.
IMHO the lack of flash is, to me at least, pretty damning. The claim is that the flash was what scared the wolf off - but as someone who takes a lot of photographs, there is no way that a flash was used in that shot, especially one from a point and shoot camera.
The thing is, these people are only separated from me by one degree - they are friends with a good friend of mine - and I confirmed the photo with the person who took it via email (and got the full-size version at that time as well).
I think it is very, very unlikely that the photo is faked. If that is the case, however, that makes it an interesting photo not just because of the subject matter, but because it is capable of making most people think it is fake even though it is not.
Secondly, if you look under the arm of the woman in the foreground, you can see a reflection from a flash there as well, probably from a piece of metal.
The photo's been looked at by a lot of people but I have yet to see a criticism that enters the realm of the scientific. The best criticisms revolve around questioning whether or not it was a wolf, which I think is a much better line of criticism, since there is no way to be sure (but it does seem unlikely).
A very good friend of mine once told me a story about how her brother's partner visited Paris, was drugged, and woke up in an alley feeling very unwell. He was taken to hospital, were it was found one of his kidneys had been expertly removed. I was dubious about it, but she told me she had seen the scar!
When I retold the story later, people laughed in my face. I guess I don't need to link to Snopes.
I'm just saying, one or two embellishments along the way can make all the difference in credibility...
I'm going to call fake on it. I've shot a lot of cameras, point and shoots included, and if anything point and shoot flashes are the easiest to spot:
- Extreme brightness and blown highlights.
- Very flat image due to the fact that the flash is in-line with axis of lens, very few shadows, and what shadows you see is very hard.
That picture is very, very much an ambient light picture - there would be the tell-tale "zombie" skin tone from a point and shoot at that range, and the people would be lit up like lanterns, washing the relatively dark background into unrecognizability.
It would take a skilled photographer with an off-camera flash to pull off that shot without destroying the background, no way that was done with an ordinary point and shoot.
When you are outside at any distance from the target, particularly if it's dusky out (which it appears to be in this picture), the flash does jack-all except for reflect off shiny things like eyes. That's just from my personal experience, having taken photos under similar circumstances and not being able to tell from the photo that there was a flash, even though I saw it go off.
It seems odd to both fake the glowing eyes and edit the EXIF. Why not just give the wolf normal eyes? And why reduce the credibility of the fake by claiming that the flash went off in a photo with no visible flash effects? I just can't see how faking the EXIF is at all plausible in this case.
edit: From a reddit comment:
> Notice the focal length. The camera was set to its maximum zoom, and probably has an underpowered flash. Other than specular reflections off a piece of plastic (purse?) under the woman's arm and from the creature's eyes, there wouldn't be any visible effect from the flash.
That is a pretty scary picture, if it is real. ;) I second the motion that this would be great for analysis by an expert. Why don't you try submitting it to him? He might have a contact somewhere on the site.
I can't stop seeing a small black bear tip-toeing along. The resolution is too chunky to say yay or nay to a photoshop on the basis of the body of the beast but I would agree with the other comment that the eyes do seem to be quite luminous considering the distance and the rest of the flash. Such a flash surely should have made the colours of the people a lot brighter and the woman's face, much much closer to the flash, is basically in shadow.
If the picture is indeed not faked, a bear would make more sense; to my understanding, wolves generally avoid humans and are not curious in the way a bear cub might be. It's a better fit for the color and fluffiness, too.
It's definitely too big to be a wolf. The size would be appropriate for a bear, and bears are probably more likely than wolves anyway. It's not quite the right shape for a bear, but that's just my opinion.
Doesn't appear to show any manipulation, though I really haven't looked at the implementation or paper to closely to know what I should be looking for.
I'm a gambling man (as my profession). I will offer you (3:2) - my $3 to your $2 says this photo is real. Let me know how much you want to bet, I'd accept someone reputable on HN will escrow. Would you trust the author of this article as the arbiter of the wager, using similar techniques as the Victoria's secret photo?
[I have no expert knowledge on photos, wolves, or the photographer in question. But I like a good bet, especially in the pursuit of science!)
How will you define "real"? Clearly the representative set of bits that make up the image are not "real". If you're attempting to determine whether the underlying scene was real then this photo alone may not be the best evidence. If you're trying to determine whether the image is original and unmodified, then you have already lost. The image is modified when the picture is saved on the camera itself and then any time the image is saved in any other program.
If you clarify your definition of "real" then your bet may become more enticing.
ErrantX said "I bet there's been some tampering going on". That's what I was referring too. As is the theme of this thread. I'm not getting in to the nitty definitions of the what "real" or "tampered" means because I think we all know what is being discussed here.
If it weren't for the eyes, I'd be more likely to believe it. The flash from a p&s just wouldn't be bright enough at that distance to generate the luminance in the eyes.
Nocturnal hunting animals have much more reflections in their eyes than humans. Try photographing cats, they can produce the same kind of luminance in the eyes.
i would bet that it is real, if someone was trying to fake it in photoshop, they would make it look overly "realistic" by dimming the eye glow and adding a shadow and stuff...
the fact that we're arguing that the wolf lacks a shadow and has overly bright eyes might be a hint that it is real simply due to the fact that an original photoshop artist probably wouldn't include all those "abnormalities"
The two pictures the author showed after saying, "Here are two photos of the same model" are of two different women. I'm pretty sure about this, because I know who the first woman is. So it appears he did catch them on some things, but he's made at least one mistake himself. Maybe someone else will know if this led to any other mistakes on his part?
That would be surprising because the woman look very alike. But if you are correct I bet VS was betting that people would think they are the same women. Meh. Doesn't really matter.
When I read that he used PCA to detect that they modified her bust, I really couldn't believe it. That is really impressive.
edit: I guess I should add a little background. PCA was used a good bit in computer vision (and data mining in general) when I was in school. You typically don't see it mentioned outside of those contexts.
I'm finding it hard to make the jump from the description on the wiki page to how you get something like the 1st Principle Component image he has there. Can anyone explain further?
It isn't about criticism, it is about forensics, using reasoning and computer algorithms to figure out what people did. It is an intellectual challenge, not a criticism.
"The word critic comes from the Greek κριτικός (kritikós), "able to discern" , which in turn derives from the word κριτής (krités), meaning a person who offers reasoned judgment or analysis, value judgment, interpretation, or observation . ...
Besides the very interesting analysis, is anyone besides me disturbed by her face in the various analysis images? It looks really, really disturbing in some pictures.
This was used as an example in a short lecture about JPEG compression in one of my classes this term. Here are the slides if you're interested. They give a basic overview of perceptual coding and the 2D Discrete Cosine Transform.
If the "ELA" indicates edited regions, why is the region where the most obvious change happened (the bag) one of the darkest?
I'm also not convinced on the lightening of the skin tone. The lighting is entirely different in both pictures, and I have friends who have shifted tone more than that after a summer of tanning. (edit: her hair is much darker as well)
This is enlightening! I'm an amateur photographer who likes to post-process some images (mostly tweaking contrast, brightness and the hues). This post taught me quite a bit on photography forensics.
Is it just me, or does anyone else feel like it would have been much more worthwhile to use an example that actually might have been 'tampered' with, and not edited in post-production? Geeks and their VS--wonder what he did with the lingerie catalog...
'Error level analysis (ELA) works by intentionally resaving the image at a known error rate, such as 95%, and then computing the difference between the images. If there is virtually no change, then the cell has reached its local minima for error at that quality level. However, if there is a large amount of change, then the pixels are not at their local minima and are effectively “original”.'
From http://www.hackerfactor.com/papers/bh-usa-07-krawetz-wp.pdf, a lot of interesting stuff in there.