Jump to content

  • Free consultations and support
  • Live chatClick Here for Live Chat
  • Call ico 1888-906-1888
    Phone support: Open

    Ready for your call :)

    Our business hours:

    Mon — Fri, 2am — 8pm (EST)

    US & EU support teams

    Phone support: Closed

    We are back in: 1h 20m

    Our business hours:

    Mon — Fri, 2am — 8pm (EST)

    US & EU support teams

Detecting Faked Photographs Gets Easier?

  • Please log in to reply


  • Guests

Posted 25 July 2004 - 10:21 PM

I came across an article about a research of Dartmouth's professor Hany Farid. This research detects whether any modifications were made with a given image analysing statistics of the pixels.

[quote name='http://www.dartmouth.edu/~news/releases/2004/07/01.html']Farid and Dartmouth graduate student Alin Popescu have developed a mathematical technique to tell the difference between a "real" image and one that's been fiddled with. Consider a photo of two competing CEOs talking over a document labeled "confidential - merger," or a photo of Saddam Hussein shaking hands with Osama bin Laden. The Dartmouth algorithm, presented recently at the 6th International Workshop on Information Hiding, in Toronto, Canada, can determine if someone has manipulated the photos, like blending two photos into one, or adding or taking away objects or people in an image. [/quote]

Professor believes that his method will eventually be recognized even in a court of law. For some reason I am sceptical about the whole concept. Most probably the reason is that I cannot think of a way you can stretch a pixel :)

[quote]A digital image is a collection of pixels or dots, and each pixel contains numbers that correspond to a color or brightness value. When marrying two images to make one convincing composite, you have to alter pixels. They have to be stretched, shaded, twisted, and otherwise changed. The end result is, more often than not, a realistic, believable image. [/quote]

#2 Qbee


    Apprentice Designer

  • Designer
  • 35 posts

Posted 26 July 2004 - 10:02 AM

can't images be shaded and stuff to make a pic look better rather than directly going for the idea that it's been manipulated.. ?

#3 mastermesh


    Apprentice Designer

  • Designer
  • 20 posts

Posted 16 February 2005 - 02:56 PM

The problem with this is that natural images are not always realistic looking. In other words, you could probably take digital photos of certain things, not alter them, and they'd come up in the algorithm as being digitally manipulated. Scary since they are going to base real world security on this stuff... Just think some day you might be arrested because you sold a photo to a jounalist in its raw format and the NSA thinks you altered it to make it look like the President is having relations with that intern when in fact it's real!

#4 Darkmonkey


    Apprentice Designer

  • Designer
  • 40 posts

Posted 16 February 2005 - 06:40 PM

can't images be shaded and stuff to make a pic look better rather than directly going for the idea that it's been manipulated.. ?

Well, if then, it's still being manipulated. Manipulating something doen't have to be wrong-doing. It's just changing something.
...they're watching you...
Darkmonkey Designs

#5 pentool


    Elite Designer

  • Designer
  • 264 posts

Posted 21 February 2005 - 02:32 AM

I wonder what image formats this applies to? Since JPEG already trashes pixels from an image, you not going to find the original pixels in the image any longer. If you re-save an already compressed JPEG yet again, the algorhythm will modify - yet again - the image pixels. So how far can you go with this, and then tell someone "this is not the real image"...?

#6 digitalmunky


    Elite Designer

  • Designer
  • 268 posts

Posted 15 April 2005 - 01:01 AM

The algorithm checks for differences by comparing pixels within the image.... It's not just looking to see if it's a digital image with say artifacting... Also, the tool would be more for aid in direction of questioning. Think of it this way, if you took a photo, whether digital or film (and then scanned), the algorithm would search for obvious manipulations, not just artifacting... It would search for variations in lighting (eg. man on left has light source on left side man on right has light source from above [yes, an obvious difference I know]). It would be the obvious (to the algorithm, not to the naked eye) that there were manipulations that could lead to ask the question WHY were these manipulations made... Something like this would less than likely be used to prove that little Johnnie manipulated a photo he found on the Net for his digital art class... As for "real" image, unless everyone that you are trying to prove your photo is real to is present when the photo is taken you will always have doubt (even if they were present, it's possible not all would believe what they saw)... Just things to think about.

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users