Re: Image forensics

[Imri Goldberg <lorgandon () gmail com>]

On Mon, Dec 28, 2009 at 11:04 AM, Martin Tomasek <tomasek () ufe cz> wrote:

> I wonder what would image manipulation detectors detect in photos with
> heavy tonality corrections, such as photos from inca's gold exhibition
> in my gallery:
> http://www.ufe.cz/~tomasek/gallery/zlato_inku/index.html<http://www.ufe.cz/%7Etomasek/gallery/zlato_inku/index.html>
> <http://www.ufe.cz/%7Etomasek/gallery/zlato_inku/index.html>
>
> or in photos with split toning:
>
> http://www.ufe.cz/~tomasek/gallery/zvire_v_pasti/content/DSC_2822_large.html<http://www.ufe.cz/%7Etomasek/gallery/zvire_v_pasti/content/DSC_2822_large.html>
> <
> http://www.ufe.cz/%7Etomasek/gallery/zvire_v_pasti/content/DSC_2822_large.html
> >

This reminds me the subject of automatic forgery detection, that was
discussed during the last Israel-Lebanon war.
At that time, John Graham-Cummings' algorithm (
http://www.jgc.org/blog/2008/02/tonight-im-going-to-write-myself-aston.html)
was used to detect forgery in photos from that war (
http://lemonodor.com/archives/2008/02/protecting_journalistic_integrity_algorithmically.html).

I tried to brainstorm a little bit with a friend on what other automatic
methods could be used to detect various modifications. Since seam-carving
was all the rage back then, we tried to think what would be a good algorithm
to detect seam carving when used to remove parts of an image. The idea we
had was that since seam carving removes parts of the image with low
gradients, you could look for parts of the image that don't have low
gradients, even though "they should". We never put this one to the test, but
more ideas to automate photo forgery detection are always interesting.

Cheers,
Imri


-- 
Imri Goldberg
--------------------------------------
http://plnnr.com/ - automatic trip planning
http://www.algorithm.co.il/blogs/
--------------------------------------
-- insert signature here ----

_______________________________________________
Fun and Misc security discussion for OT posts.
https://linuxbox.org/cgi-bin/mailman/listinfo/funsec
Note: funsec is a public and open mailing list.