Video: Fauxtography made easy!
posted at 8:55 am on September 4, 2007 by Allahpundit
Send to a Friend |
Share on Facebook | printer-friendly
Brilliant. There’s nothing here that a skilled photoshopper couldn’t do with painstaking effort, but that’s just the point. Once this widget gets picked up by Adobe for CS4, you won’t need skills or effort to pull it off. They start off with the pedestrian stuff and pick up steam towards the end, so stick with it. The ease and seamlessness with which people can be erased from a shot is especially fascinating and troublesome.
Image hoax-detection software already exists and is used (in theory) by some major news orgs, but once this new app becomes popularized it’ll be as mandatory for online news consumers as anti-virus software. Adnan Hajj, call your office.
You must be logged in to post a comment.
Blowback
Note from Hot Air management: This section is for comments from Hot Air's community of registered readers. Please don't assume that Hot Air management agrees with or otherwise endorses any particular comment just because we let it stand. A reminder: Anyone who fails to comply with our terms of use may lose their posting privilege.
Trackbacks/Pings
Trackback URL
Comments
Holy crap…even I could indulge in fauxtography with tools like that.
flipflop on September 4, 2007 at 9:03 AM
…So who you gonna believe? Me or your lyin’ eyes?
bofh on September 4, 2007 at 9:05 AM
My little brain can’t fathom the quadrillion ways this technology could be used by powerful intities to fool and misdirect the masses.
My question to it’s authors is; Can one detect it’s use as having been employed on a finished digital picture. (Don’t know if that’s a correct sentence)
Ernest on September 4, 2007 at 9:16 AM
Doooood! The ministry of dis-information has your number.
That was an awesome demo! I mainly work in vector, but I got a couple of G.A.s that have to have this widget. Thanks for the head up.
PS: How come this youtube stuff has no sound from the hotair location? I can C&P to youtube and all works fine. But how come this you tube stuff has no sound from the hotair location?
locomotivebreath1901 on September 4, 2007 at 9:17 AM
Better still, did they build in a method of detection?
JiangxiDad on September 4, 2007 at 9:24 AM
I don’t run anti-virus software. And I run Windows XP.
It’s almost unnecessary. And rarely of much use when you do have an infection.
The ones that spread are usually new viruses that the software as it exists doesn’t have definitions for. Heuristic analysis is rarely effective.
Further, with Firefox and NoScript enabled, plus a good hardware (or software — I use both and Windows Firewall is fine) enabled, I haven’t had an infection of any type for years.
Yes, you should scan downloaded files (ClamWin is good for this), but the background anti-virus is unnecessary.
Anti-AP† software?
Essential.
† Associated Press
Christoph on September 4, 2007 at 9:25 AM
Stalin would be proud…
JetBoy on September 4, 2007 at 9:26 AM
“In theory” is right, Allah. “In theory” the “major news orgs” are also unbiased and reports only the facts.
Troy Rasmussen on September 4, 2007 at 9:28 AM
Stalin would have loved this program.
Clark1 on September 4, 2007 at 9:37 AM
That’s very cool, in an “objective reality doesn’t exist anyway, so why not alter it” postmodern sort of way.
A complementary, but more frightening algorithm was publicized earlier this year: Attention Actors: Meet Your New Digital Replacements
wordwarp on September 4, 2007 at 9:46 AM
That’s quite impressive. At least we can rest assured that no news organizations would be willing to doctored digital images to project a false reality. Oh wait, nevermind.
Spirit of 1776 on September 4, 2007 at 9:58 AM
If there was any doubt before, I think it has now been confirmed - seeing is not necessarily believing.
nailinmyeye on September 4, 2007 at 10:21 AM
NOW how many words is a picture worth?
smellthecoffee on September 4, 2007 at 10:33 AM
WAAAAY cool! I gotta have it.
Warner Todd Huston on September 4, 2007 at 10:47 AM
Actually I have a question: Can this kind of technology be applied to digital video as well?
JetBoy on September 4, 2007 at 10:49 AM
Combine this with the UPI “story” on the al-Doura power plant and you have a major problem trying to determine if events actually happened or not.
Tom Blogical on September 4, 2007 at 10:52 AM
Still a thousand. They just happen to be worthless Chomskyite ramblings now.
Blacklake on September 4, 2007 at 11:19 AM
That’s true, and it works. It’s what they use to auth images used in court cases.
triple on September 4, 2007 at 11:22 AM
Digital video, like film, is still just a sequence of still frames. It’s trickier to manipulate image sequences convincingly as a certain degree of uniformity is required from frame to frame lest we detect the inconsistencies, but in this case I expect it’d be relatively straightforward relative to some of the other manipulations that can already be applied. I don’t know if it could be used convincingly for dynamically resizing, as with the web content still images featured in the video (though maybe), but as a post-processing technique I’d say it’s almost certainly applicable.
I don’t think it’d bring any new capability in that arena, though. It’d just provide some shortcuts.
Blacklake on September 4, 2007 at 11:25 AM
Boooooo on Adobe takeover. …Rubbish, filth, slime, muck…booo…
Adobe is swiftly becoming the Microsoft of graphic software. New versions have minute changes yet cost a fortune and are even more bloated than the last release. (trust me - I use them every day…)
We need some good olde free market competition!
whatthecrap on September 4, 2007 at 11:29 AM
Thanks. I didn’t know if the tech exists to perhaps manipulate one frame in a video, and have it carry throughout the clip. Such as shown here…where an image is changed through negative space so to speak. Couldn’t some software just carry out the same (let’s say) cropping, and apply it to each frame in a film?
meh…either way, the easier it gets to manipulate photos, the more of them I’m sure we’ll see. Or think we see….
JetBoy on September 4, 2007 at 12:36 PM
CS to CS2 was crap. The video management upgrades in the CS3 suite are nice, but it does suck that upgrade licensing has become such a racket.
Next version should reanimate lifeless matter.
The Race Card on September 4, 2007 at 1:14 PM
That video showed some very noticable artifacts on the city scene after the woman was removed. I didn’t catch anything on the beach seen on my first time viewing. It looks like they have some tuning to left to do on the algorithm to fix up gradients and other things.
The best part is there wasn’t anything there that wouldn’t work for video either. Great for removing unwanted relations from home movies, or jihadis from among a crowd of school kids.
pedestrian on September 4, 2007 at 1:24 PM
CNN would really like something that could on the fly edit video to change the race of perpetrators to be more representative of America, or to replace the uniforms on attackers to show us the larger truth of God’s Warriors.
pedestrian on September 4, 2007 at 1:32 PM
That’s hard to say and would depend on a lot of things: The original picture, how it was changed, whether either the original or the final picture were compressed, etc. I’d venture to guess that it’s not always detectable but often could be. For example, if objects are removed by collapsing followed by stretching, the stretching would cause certain high frequencies to be missing in the low energy paths. While low energy areas generally lack such high frequencies, the need for a path might make this possible. For example, consider the image on the beach. Even the lowest energy portion of the sand will have high frequencies. If I suspected that something were missing in that photograph, I’d analyze the high frequency portions of the sand and find that areas near the people would have higher frequencies than areas away from them, indicating some type of manipulation. Or, if the grains of sand are large enough, some might appear stretched (although perhaps not by enough to detect).
Anyway, it’s harder to check this than to check work by some amateur hack stringer with today’s Photoshop.
calbear on September 4, 2007 at 3:20 PM
I’d say once this thing get’s incorporated in a program such as After Effects such as. (Sorry had too) Anyone one like After Effects that can track an effect from one frame to the other, making sure that the same type of pixels in an area are removed across all frames then bingo. But it won’t be done on the fly like this. It will have to be rendered. And re-rendered after locking ‘key pixels’ to prevent weirdness from the first render.
- The Cat
P.S. When do they start using this on Hillary’s beach photos? The thighs man the thighs!
MirCat on September 4, 2007 at 3:39 PM
The MSM is going to LOVE this. Photoshopmania.
RightWinged on September 4, 2007 at 5:26 PM
Geeez, folks, settle down.
It’s interesting, and it’s new. But if you look really closely at the demos, you can still see where the cuts are. For instance, when the people are removed from the beach, there’s no way to create unique image behind them, so the sides are just–in essence–cloned into the void.
I kind of doubt this’ll be in CS4 either. Adobe isn’t really into innovation these days.
Typhoon on September 4, 2007 at 6:23 PM
My question would be,, if one were to take a fresh digital image of the altered image,, would that make it harder to spot?? Well,, we already have a mainstream media that lies and distorts and is ready and willing to destroy any conservative,, we have the 24 hour news cycle hungry for fresh raw meat,, pictures and images have a massive effect on the news and on us as a society,,, is it that much of a leap to expect made up photos or distorted and altered images to be passed off as fact! How many other verbal myths are still repeated as fact every week in this media.
JellyToast on September 4, 2007 at 7:51 PM
But it isn’t cloned. It’s stretched in those places where stretching would be least noticeable. Thus, although often detectable, I doubt current hoax-detection software would be able to detect it in most instances.
calbear on September 4, 2007 at 7:55 PM
Digital “photography”, since there is no analog negative produced - merely file-stored ones and zeroes- has no scientific and thus no legal basis for our trust.
All of these digital red light cameras, etc., should be challenged in court as unverifiable, since how can anyone every prove (without a time-stamped, material negative) that the image wasn’t hoaxed? (There’s no “grain”, or mechanical camera “artifacts” to signify when, where, or how the picture was taken, and if any after-the-fact manipulation was done to it.)
I’m surprised no one has challenged this illegitimate medium yet.
Especially when they want to take you money or license away.
I’m a “follow the rules driver”, so I’m not going to be hit by this, but someone needs to contest the invalidity of the digitial image in its entirety, and down to its illusory root.
Same with news photography.
No negative, no proof of reality.
The ease of the method has lulled everyone.
But, with the fauxtography scandals now bubbling up -and, as potentially-Luddite as it might seem- it is time to demand an analog basis to these “news” claims. A microfilm and digital dual-camera for news reporting, fine-demanding images from traffic cameras, etc. should be the law.
If this danger isn’t focused now, we lose one more contact with verifiable truth.
profitsbeard on September 4, 2007 at 8:29 PM
In that case, you could never prove reality. How do we know that a negative was produced by being exposed to a real scene rather than to a digital display? We can’t if the display and capture is good enough. Sorry, but anything that’s easily faked in digital is easily reproduced as analog. If you say that digital has to be invalid, so does any analog that hasn’t been verified to have been created prior to the point at which we had to dismiss digital.
calbear on September 4, 2007 at 9:16 PM
Sorry guys, but it is not possible to draw accurate conclusions about the quality of the altered images by watching the video. The video itself is a compressed form of the image, and therefore has altered the altered image yet again, if you get my drift.
-
Since some of the commenters can see traces of the alterations, it may be that the video codec algorithm has modified the image in such a way that artifacts from this technique become visible. If that is the case, the originally modified image may look perfectly convincing while the compressed video image reveals the alteration. Again, if this is true, it would be a possible way to detect the changes.
-
In any case, it should be possible to detect altations by doing some form of pixel analysis. When digital cameras save a picture to memory, that picture is processed in the camera using a number of proprietary image adjustment algorithms (only RAW images are left untouched). For example, most cameras have an image sharpening function that enhances edge contrast. The modified pixels in the altered image would most likely not be consistent with the output from this sharpening function, which would suggest alteration.
Herikutsu on September 4, 2007 at 10:27 PM
When you have an artficial light source there are going to be gradients cause by increasing distance from the source and also the changing angle of reflection. When they remove those slices that rate of change is going to be affected. You can see in the foreground below where the woman was removed a curved edge of increased gradient. It looks like a sharp bump in the road, which makes no sense in a city scene.
I think if you really wanted an artificial background to replace a removed figure you would use some type of pseudo-random pattern that was adjusted to match the histogram and spatial frequencies of the surrounding area.
pedestrian on September 4, 2007 at 11:01 PM
“Hooray for Pallywood!”
hadsil on September 5, 2007 at 3:33 AM