How can you tell if a digital picture has been altered?

This photo has obviously been tampered with:

http://www.geocities.com/wearetheromansca/grassykb.jpg

However, if end effect was something not so obvious, I wouldn’t have seen it. Is there any telltale signs that a image has been altered?

Hmm… your link doesn’t seem to wanna work.

Good digital alterations are damned hard to detect. I work with Photoshop extensively, and I know just how easy it is to cut and paste. But there certainly is an art to doing it effectively and convincingly.

One thing people screw up on is inconsistent lighting. The light on whatever’s been spliced doesn’t match the lighting of the overall scene, even when artificial sources are taken into consideration. Another thing is focal length. Sometimes you will have a picture where something two inches into the field is in focus with something at infinity. Even at f/22 or f/32, most lenses cannot do this, especially at the 50mm or higher range.
A fisheye, yes. A telephoto, most definitely not.

Also, look for shadows that aren’t where they’re supposed to be, look for bad aliasing around the edges (these are smoothed out sections that appear between something that clipped and the background it’s pasted onto.)

What else? Objects appearing in incorrect perspective. If you cut a photo that’s taken with an 85 and composite it on a photo taken with a 24, you will usually notice that something just ain’t right.

Repetitions of pixels or patterns in the picture that shouldn’t be there is another sign of doctoring. I don’t know. There’s a hell of a lot of things you can look for. The best thing is to ask an experienced photographer her/his opinion. There are a lot of images that are physically impossible to get in still photography. I’ve only covered a few cases.

Random points:

You have to hit reload after following the link. Geocities doesn’t allow outside links directly to pictures.

Also, the name of that doctored picture is “In-A-Gadda-Da-Oswald”. Here’s the original source: http://www.tw-zone.com/cosmo/photoshop/oswald.html

Study it closely, if you’re not sure. Look for differences in grain texture, too.

This link should work. http://www.geocities.com/wearetheromansca/grassykb.jpg?

The funny thing I’ve noticed is that when shown a strange picture, many people will say something to the effect “that’s the worst photo editing I’ve ever seen!” even when the pic turns out to be real.

An example is a recent thread where a pic of some Brit actress who’s septum had dissolved away from cocaine use. I’d search for it, but I don’t quite care that much.

A little informational note: I showed the pic in the OP to a friend (because it’s funny), and he noticed the “Dead Kennedys” emblem on the back wall, something I failed to see. I’ll bet this is from a Dead Kennedys album, perhaps album cover art.

Inconsistent color casts and contrast levels are often clues. As lissener said, differences in graininess across the image are a dead giveaway.

The trouble isn’t the major hack jobs - they’re pretty easy to spot. It’s the subtle “enhancements” made to straight photos. Spot out a blemish here, rub out a distracting background element there, add grandma into the family photo because she would have been at the reunion, if she hadn’t died last year… Regardless of whether digital or traditional methods are used, when does a “photo” become an “illustration”?

(Personally, I touch-up every photo I scan to some degree. But I gather that what OP means by “altered” is adding or removing image elements.)

Ok, it’s not such a big deal in the art and advertising fields, but things get kind of dodgy (excuse the pun) when a touched-up photo is used for news reporting. Remember the O.J. Simpson mug shot photo on the covers of Time and USA Today (IIRC)? Both were the same original photo, but one of the magazines darkened O.J.'s beard to give him a 5:00 shadow. Made him look way sinister. Controversy erupted, and the mag’s response was that since it was the cover, it was advertising, not editorial. Hence, ok to do O.J. Personally, I think they stepped over the line for a news magazine. If that was Cosmo, it would have been fine.

(Not that I’m trying to hijack this into an O.J. discussion, 'cause I most certainly am not. I couldn’t give a crap if he’s guilty or not. It’s just the best example of my point I could think of.)

Just last week a friend showed me an article in a British car magazine about a new Jaguar convertible. If the captions hadn’t said so, we never would have guessed that the photos were digitally created. (The car, in fact, hasn’t even been built yet.) These were full page, full color images of a car driving around outside in full sun. We examined them for half an hour - we’re both photo geeks. Couldn’t find one obvious clue. Frightening.

I’d post a link, but I don’t think they have a web presence. And I forgot the name of the magazine. Doh!

pulykamell has it right, though: anybody can cut and paste, but it still takes talent and skill to do it convincingly.

I was just informed that the car magazine was Autocar dated 14 March 2001.

Kamandi – Autocar, really? I used to take photos for 'em, until I got sick of the way they treat freelancers. I’ll have to check that magazine out. I hate it when magazines do that, and don’t make it too terribly obvious that it’s a computer jobbie. In a magazine, there’s a lot of leeway, but I still think it’s unfair for the reader. As a photojournalist, I both love and hate Photoshop. It’s a wonderful tool, but it seems to me that people have been misusing it, and the public has gotten generally skeptical about photography. You’ve mentioned the OJ thing, another issue was, what’s her face, the woman in Iowa who had the octoplets or nonopulets or whatever. Anyhow, on the cover of Time, a portrait of her was doctored to fill in the chips or gaps in her teeth. This was not done by anyone in the photo dept, but by someone in production. Photojournalists were PISSED. This is a big no-no in the news world. Doctoring a news photo like that for aesthetic purposes is absolutely prohibited. Then there was the issue of National Geographic on Egypt. The photographer took a photo of camels and pyramids in the background, but in order for the photo to work on the cover, the pyramid on the left had to be digitally moved a few centimeters to the right. Once again, photojournalists around the world were none too happy, and National Geographic fessed up to its fuck-up.

BTW, the OP’s photo (I finally loaded it) is pretty hilarious. The only dead give-aways on a quick look were the Dead Kennedys symbol in the background (to me it looks obviously painted in) and the Roland keyboard’s lighting is a bit off. But not too noticibly. Whoever did that did a pretty damned good job.

      • First off: I had to use mblackwell’s link and erase the question mark off the end to get to the image.
  • Usually in an altered digital photograph, you can find rather abrupt differences in the histograms of altered and non-altered places (-or suspected altered and non-altered, I guess…). Color images are harder to alter because in a b/w picture you only have to contend with the grayscale but with color images you have three color channels to get right also. How the image has been encoded is important too; jpg is a lossy compression, and robs the image of much of its grayscale accuracy and image detail. Bitmaps are easier to find differences in because there’s no loss in compression. -And in digital images, bigger is better, for testing purposes.
  • A histogram is a display of the distribution of various aspects of the color of the image. You can find out the intensity and distribution of colors in the whole image, or in any selected spot (Paint Shop Pro has this feature). You can add fake stuff to a digital photo pretty easily and get it to “look” right, but it’s damned hard to get the histogram of the added object to match the histogram of the background, because the computer can detect differences your eyes can’t see.
  • This is because the image is a record of the relative intensities of colors in the lighting in which the original image was taken. So for instance, if you have a person sitting in a shady spot, there’s a limit to how bright or dark that any color in that original photo will be, that is dependent on those lighting circumstances. This limit depends on all of the lighting captured in the photo: so a picture of a statue outdoors taken on a sunny clear day will have a different histogram from the same photo taken during an overcast day, even though both will have the same colors. ~ Finding a different histogram for a suspect object in an image is about as close to “proof of alteration” that you can get. - MC

Another thing to look for is unnatural-looking discontinuities in one particular color channel which appear normal when viewed in RGB. Often times, you can touch up a color photo to the point that it looks great overall, but if you view just the red channel, for example, you’ll see obvious brush marks. This happens a lot in darker areas of a photo, where variations in one channel are difficult to notice when mixed in with the other channels.

Whenever I’m doing a hack job, after I think I’m done, I go to each channel individually and find that I have much more smoothing to do, and it usually helps the results in the overall image, too.

      • Just poking around in the Oswald photo, it’s very hard to find anything out of line because the detail is so poor when the image is magnified (jpg’s have tell-tale differences in graininess simply as a result of compression). One easy givaway is that the fake shadows that have been added (such as under the keyboards and behind the microphone) only seem to have 3 different grey levels, where the original shadows have a whole bunch (12-18) of different levels. Another givaway is that the keyboard keys and Roland lettering have a white level much higher than anything else in the photo. - MC

Oh, duh! This is an obvious error-- Back to the keyboard again. It’s soft. I mean, it’s a bit out-of-focus, when it’s CLEARLY within the depth of the field. Check it out. The guitar is in focus. The keyboard is out. The people in the backgroun are in. There’s no way to achieve that kind of effect with photography.

That’s the dead give-away. Otherwise, the Roland’s a bit light for the photo. The people are wearing (I assume) grey suits, and Roland keyboards are dead black, so they should be much darker, even if hit with a flash.

Also, the hand where Oswald’s gripping the mike looks like it’s been smeared a bit, or something. Also, Oswald’s right hand seems to be going through the mike stand, or something. I can’t quite tell what’s going on there, it just looks odd. Also, the keyboard seems to exist in a physical space where there isn’t enough space for a keyboard to be. Compare Oswald’s position with the “keyboardist”. Look at the legs. And, again, the perspective lines of the keyboard don’t seem to match up with the rest of the photo.

Oh, duh! Geez, how did I miss this first time around. There is one dead giveaway, irrefutible proof of the alteration. Ah…-- Back to the keyboard again. It’s soft. I mean, it’s a bit out-of-focus, when it’s CLEARLY within the depth of the field. Check it out. The guitar is in focus. The keyboard is out. The people in the backgroun are in. There’s no way to achieve that kind of effect with photography.

That’s the dead give-away. Otherwise, the Roland’s a bit light for the photo. The people are wearing (I assume) grey suits, and Roland keyboards are dead black, so they should be much darker, even if hit with a flash.

Also, the hand where Oswald’s gripping the mike looks like it’s been smeared a bit, or something. Also, Oswald’s right hand seems to be going through the mike stand, or something. I can’t quite tell what’s going on there, it just looks odd. Also, the keyboard seems to exist in a physical space where there isn’t enough space for a keyboard to be. Compare Oswald’s position with the “keyboardist”. Look at the legs. And, again, the perspective lines of the keyboard don’t seem to match up with the rest of the photo.

Is this a digital cut-and-paste job, or a scan of manual cut-and-paste job? I recall seeing the same illustration about 10 years ago.

labradorian, you could’ve seen it 10 years ago and it could STILL have been a digital edit job.

::launches vMac with System 6.0.8::
::launches Photoshop 1.0::

Yep, copyright 1989-90. That pic could’ve been altered on a IIx a dozen years ago.