A Tool to Detect Image Manipulations?

I found this site. errorlevelanalysis.com... You can link images for an Error level analysis.

Error level analysis shows differing error levels throughout this image, strongly suggesting some form of digital manipulation. Areas to note are the lips and shirt, as well as the eyes. All are at significantly different error levels than their surroundings. Presumably, colours have been altered and areas brightened.

Check out the link for the said image. Here is an example I just used in another thread.

errorlevelanalysis.com...

I believe that shows a fair amount of manipulation. I also believe this could be the result of simple enhancements, I'm not an expert. Maybe someone with more knowledge can comment.

I did This pic. it's of a UFO. Clearly fake. But I don't know how this site works??

it looks like it just helps discern pixelation of certain pictures...anything that looks highlighted with more of a reddish color seems to indicate that it's been photoshopped...you're picture of the x-37 does not look photoshopped (if it is they did a very good job) as for the UFO there are changes in pixelation that are pretty evident even in the sky...I don't personally know if it can help discern certain photoshop jobs on pictures i think it just helps pic out whether there are any enhancements or changes to an original picture...the UFO picture could have easilly just been a copy paste of a lamp or something

A good tool to have to detect manipulations. I believe the link you provided does not go into enough detail as how to determine a manipulation. I would think perhaps one has to purchase software to get more details and work with images. I do digital enhancements all the time - not to hoax or anything just to change my images to make them more interesting in an artistic way, I post them on my online art group. I can usually tell when an image has been manipulated or enhanced. The UFO spaceship would have been layered over the top of a background. S @ F

Originally posted by here4awhile
it looks like it just helps discern pixelation of certain pictures...anything that looks highlighted with more of a reddish color seems to indicate that it's been photoshopped...you're picture of the x-37 does not look photoshopped (if it is they did a very good job) as for the UFO there are changes in pixelation that are pretty evident even in the sky...I don't personally know if it can help discern certain photoshop jobs on pictures i think it just helps pic out whether there are any enhancements or changes to an original picture...the UFO picture could have easilly just been a copy paste of a lamp or something

edit on 1-1-2011 by here4awhile because: (no reason given)

Ahh... so generally, it is useful only to the extent of photoshop: i.e. Useless.

I still think the image I linked in the OP was edited because I tried an unedited photo and it was mostly black, no color.

The best method for detecting image manipulation is based on 'Camera Pattern Noise' by:

Jan Lukáš, Jessica Fridrich, Miroslav Goljan, 2006, 'Detecting digital image forgeries using sensor pattern noise', Security, Steganography, and Watermarking of Multimedia Contents VIII

This method has a near 100% accuracy in detecting forged images. It does help when the image is at the largest resolution possible. It works by identifying the unique noise pattern of each individual camera sensor, like a fingerprint. This is then referenced to the pixels to determine if alterations have taken place. I am yet to locate any software on the open market that uses this technique. It is expected that some USA government agencies do have access to software based on these techniques. The movie 'Beyond Reasonable Doubt' with Michael Douglas shows it in action.

Tried to do a test with a straight from the camera image and 95% compression jpeg of the same image. I get this "status: Image is too large. Our little server can't process stuff that big " and it's only 4.2mb jpeg and the 95% is 2.1mb
This I don't think is a reliable tool for detecting manipulations. I think noise analysing works way better as it reveals cloning etc. I think I'll device another test just for kicks

I take that back. It does detect differences in jpeg save ratios. I tried by taking the same raw image and then on top of the original I pasted a part of the image that was saved multiple times with different jpeg compressions. The analysis clearly shows where the paste was, under the red feather of the dream catcher.
Link

Unfortunately the service this website offered no longer exists, they cite money and time inputs from volunteers as the contributing factors.
Well, i thank them for their efforts, it was a useful tool, i hope they manage to implement something similar in the future.

If anyone has links to other tools such as this, could you please post them, i would be very grateful.

They will probably make a program out of this or something. The problem with this service was that most people didn't seem to understand how it works. One member for example claimed all hurricane sandy photos were fake because he didn't understand the process.