Specifically “ghost” photographs. The easiest and earliest was to fake a ghost photo is a double exposure-and i am sure that analysis of such a photo can be easily done today.
Now, there are numerous examples of photos (alleged to show ghosts). Can these be easily analyzed by software obtainable today?
Are there any such photos that cannot be explained by obvious fakery?
Of course not. There are always things you can check for that might reveal a fake, but a clever enough fake can circumvent those. You can think of some clever new detection trick, but then the fakers can figure out how to fake that one, too. Take a (supposed) Bigfoot photo, for instance: You might be able to detect if Bigfoot was digitally composited into a real photo of a forest, but what if it wasn’t digital? What if it was a wax and hair statue of a Bigfoot, that cast all the same shadows and so on as a real one? OK, maybe you could pick up the lack of motion blur… So replace the statue with a human in a Bigfoot costume, and have the human actually moving. Now maybe you can determine from perspective cues that the supposed Bigfoot is only 5’10"… so put Ted Cassidy in the suit, instead. It’s still a fake photo… but it really does show a genuine tall hominid covered in hair. The only fake part is the claim as to the hominid’s species, and how is the detection system supposed to determine that?
There is photo software out there that helps detect fakes. Tungstene is the main one I know of. Some wire photo agencies use this software, and World Press Photo is using it to help with issues it’s had in the past when some of its winning photos were manipulated beyond standard industry accepted color and tonal adjustments. Can it detect all fakes? I have no idea what its success rate is (and can’t find any published numbers or even claims), but I doubt it’s 100%.
As for “ghost” photos, there’s plenty of examples of “ghost” photos I’ve seen out there that are not doctored in the least, but rather the result of various lens imperfections that cause haloes of light or little orbs to appear because of light diffracting and bouncing around a glass filter or lens elements in a way to leave a weird glow behind. Or just from light hitting nearby out-of-focus dust particles or raindrops or whatnot and leaving a glowing circle behind. Look up “orb photos” for all that silliness. That would not show up as a “faked” photo. It’s a real photo, just with odd imaging artifacts.
You’re mixing two different concepts: your first example is a classic ‘photoshop’–digitally adding an image of bigfoot where nothing exists. This is totally fake, and this is what I think the OP is asking about…can such fake additions be detected?
For example, would it be possible to place an electronic signature on a jpg file, which would, say, add a code number to each existing pixel? After the signature, any pixel which is changed would be given a different code number, and it would be easy to detect that somebody had photoshopped the pic.
Your second example (the wax statue) is not a fake photo–it’s a real, genuine photograph taken of a physical object (the wax statue), placed in a real forest. The photo is completely genuine, (and you can be totally honest about it , if you want to: just label it “wax model of an imaginary bigfoot”.) If you label the photo “proof that bigfoot exists”, then you are a liar…but not the photo. The photo is a a real picture, of something you saw physically in front of your lens. It’s not a composite photoshop, of something that never existed.
You can imagine a situation where the faker has access to all the fake detection systems, and is expert in their use as well. So he iterates his efforts, fixing flaw after flaw in the fake until it is undetectable. This is almost certainly not possible, as fixing some artefacts will bring about others whose fix is incompatible with the original fix needed.
There are lots of things to look for in detecting a fake that simply messing about in Photoshop will never allow you to get right. Some quite simple operations performed in frequency space can start to show up all manner of traces of fakery, some that are very subtle if viewed in the original image. Digital images carry many traces of their provenance, all the way from the particular algorithms used to process the Bayer grid, to the compression algorithms and so on. Faking a picture and avoiding leaving footmarks in these is not going to be easy without specialised software to help.
Actually, there is an easy way to get around this problem: print the photo on a high-res continuous tone printer, and then take a photo of the output. The resultant low-pass image should get rid of any digital manipulations, and if it’s blurry enough, should easily pass for a real photo of Bigfoot, or Nessie, or Casper.
yes, if you look, you will find many photographs claimed to depict “ghosts”. my question was-can modern software detect these frauds right away? second, have any such photos been found to be “not tampered with”? Many of these pictures are claimed to show something that the photographer did not see while taking them-seems a bit odd. the images only appeared when the film was developed.
Actually it may not. Depends what spatial frequencies the artefacts are in. The neat thing is that quite high amplitude artefacts can reside in images and be invisible to the eye.
Once a photo is so blurry and poor, the question of fakery becomes moot.
As a matter of technical problems, if the photo was on film, I would want to see the negative. The dynamic range of film is quite a bit wider than any printed format. And it has a very characteristic transfer function. If the neg only had the dynamic range of a printed paper there would clear evidence of fakery. Indeed I would be looking at the neg for all sorts of further evidence. Are the geometrical distortions consistent with the claimed optics used? Are the various traces of things like the film mask consistent with the exact camera used? Film cameras are like typewriters. There are traces than can identify pictures taken with them.
And of course the picture on the negative had better be on the roll in the correct sequence with the other pictures taken with the camera. And all those other pictures had better have all the right technical traces as well.
As in my post above - with film I would be looking at the negative first up. This isn’t a software question, although maybe some software assistance might help speed things.
Chemical contamination has been blamed for some ghost artefacts (no cite, clawing back into my memory here.) Also simple light leaks.
If your ghost is some indistinct blurry being that maybe reminds you of ideas of ghosts in shrouds, film presents lots of ways of getting such artefacts. When it comes to a ghost image, I’m much more interested in a bloodied torso carrying its head in one hand. Something much more compelling here.
Actually, I think it’s perfectly relevant. When presented with a purported ghost photo, the primary question is “has this photo been faked?”, that is to say, “is that a real ghost?”. Only after that question is answered is the question “how has this photo been faked?” really relevant. It matters little whether the fakery was done in a darkroom, or in a software program, or with physical objects.
Furthermore, as soon as you grant that undetectable fakes could exist created via physical props, then you also reach the conclusion that undetectable fakes could exist created in software. Consider that fake image created with props: All it is is a sequence of ones and zeroes. There’s no reason you couldn’t create a computer program that would produce that exact same sequence of ones and zeroes as output. It would probably be difficult to do so, aside from trivial examples, but it’s possible. And realistically, you don’t need to produce that same exact sequence, either: There are a great many potential images that will appear “genuine” to any detection scheme (after all, there are a great many images that really are genuine), and you just need to produce any one of them. And as the detection schemes get better, so will the techniques for producing images.
You missed the point ralph.
Many of the ghost photos are NOT photoshopped. You’re seeing exactly what the camera saw. But that differed from what the human saw because the camera lens & camera sensor are different from the ones in your eyes.
So for example the camera captures a burst of light that existed only inside the glass of its lens. The burst of light is 100% real.
What’s 100% fake is the camera operator assuming the lens flash represents a “ghost”, whatever the heck a ghost is.
With enough effort any photo or digital image can be manufactured and be indistinguishable from a record of reality. There is no guaranteed way to tell the determine that an image has been altered. ‘Real’ images are just fakes of reality anyway. Most fake altered images can be detected because it’s rarely worth the effort to hide every last artifact.
Oh, another problem with ghost pictures specifically, that doesn’t show up with Bigfoot or Nessie: Bigfoot or Nessie, if they exist, are presumably real physical objects, made up of basically ordinary matter, which interact with other matter and with the laws of physics in ways that we’re familiar with. With them, you can notice things like a lack of shadow, and say “A-ha! That can’t be real!”. But we don’t know what the rules are supposed to be for ghosts. Maybe ghosts cast a shadow, maybe they don’t, maybe they sometimes do and sometimes don’t depending on some factor we can’t even guess at. Maybe ghosts really do exist, but just happen to really look almost exactly like lens flare, or like JPEG compression artifacts, or whatever. How could we know?
Lens flare is ghosts in the same way that electricity is magic.
Spoken like a true ghost agnostic
Interesting.
Couldn’t tell if it (or a variation) is in public release.
Naturally intelligence agencies have been doing that since before electronics.
The first sentence is key. And actually pulls the rug out of the second one.
Ever photo you have ever seen in a newspaper or online has been altered. That’s what photos are…and no photographer of basically anything will not alter images not only in ways conceivably traceable to optical techniques–exposure, balance, saturation, etc.–but now digital focusing (depth of field), for example and retouching and even layers of cut and paste are available in many apps for free to a few dollars, and are being bought by everybody who takes a selfie. Which is a lot of people.
So the experience of “is that real or not” will not be brought to bear as often in the future at all.
Not as I read the OP. The OP specially asks about software to detect photo fraud not if people try to purpritrate fraud.
Yes. What I asked was: you see a photograph that shows a image of what appears to be a human-but the image is transparent and “ghostly”. can you determine if this is an actual photograph of something, or is it am image of something produced outside the camera?
It’s already been answered.
No, you can’t.