Have you ever found yourself wondering whether a certain celebrity photo was enhanced or altered? These days it may in fact be the bigger challenge to do the opposite, and actually find that magazine photo that was NOT Photoshopped.
Digital photo enhancement is endemic, but with the unprecedented computational power of modern computers, would an algorithm be able to spot the image alteration that the naked eye cannot? A team of Computer scientists in Dartmouth College in New Hampshire think so, and are doing the research that paves the way for a tool that could analyze an image and give the likelihood that it was altered on a 5 point scale.
What happens in retouched photos? Wrinkles disappear, skin blemishes clean up, several pounds of weight are slimmed off, imperfections removed; with the end result not resembling anything that exists in the real world, even as they are being sold and pushed as the ideals to aim for. The alternations, moreover, exits on a slippery slope: why stop at clearing the skin when you can also raise the eyebrow, shave off inches off the waist, etc. You could say that the quest for beauty requires this sort of thing, but is it ok if the truth is sacrificed in the process?
But imagine the following: browsing the web with a little number that appears on the corner of each image that indicates the likelihood that an image was altered (yes, without having access to the original image of course). It’s not quite here yet, but my guess is that it’s just around the corner. Prof. Hany Farid, a computer science professor at Dartmouth (shown to the right presumably in an unaltered photo 
) and PH.D student Eric Kee, are working on this research, inspired by legislation that is being pushed in Europe to attach disclosure labels to images that have been digitally altered. You can see Dr. Farid’s gallery of before and after images here.
Want to learn more? Here are some links:
- Virtual Reality: Dartmouth Researchers Challenge Photo Retouching with Rating System (on the Dartmouth College Website)
- Photoshopped or Not? A Tool to Tell (New York Times)
- Dr. Farid’s before and after image gallery (again)
Thoughts or comments? Share them in the comment section below.
It’s not here yet, but I hope in the near future such an app comes available!
Can I ask for another one: a bullshit detector for financial experts who pretend they know what they are doing when talking about the economy?
@ Toni: haha. Indeed.
That’s easy, just use the code below:
Function isBullshit {
return true;
}