Image Tamper Detection

[hbk]

i talked to a few friends who are into photography.....

they suggested that in case i hav just the 1 image, what i can dop is to check for the continuity of a number of features in the image.... coz in a natural untampered image, all features are in a cntinuum..... no drastic variations...


eg. i can check for continuity of sharpness, brightness, and other such features....


but this isnt foolproof, is it?? i mean, dpoesnt it depend on the lack of skill of the person doing the tampering.... the more skilled he is, the more difficult the detection.... wat say???

[elric]

yea, the technique your friends suggest is quite flawed, for an untampered image may also have a drastic variation, say in brightness if there are some metallic glare

[xheavenlyx]

Exactly as Elric said, and my above post. Its not possible, unless they guy is modifying the picture using Microsoft Paint.

Your best bet it to have your originals made in a way that only you know the brightness and stuff continuum (unless you are doing tamper detection on many other sources of which you have no idea of the original)

Secondly, You can "visually" try to see the changes with your own eyes! Much better if used in combination with MATLABS feature extraction etc...

[maurits]

I know two basic approaches to do this. The first method processes the suspect image region by region and looks for similarities that might have been caused by the 'cloning' tool, or coping image regions.

Have a look at the paper by J. Fridrich, 'Detection of Copy-Move Forgery in Digital Images with D. Soukal and J. Lukáš'

The other, and probably more difficult approach is the following. A digital camera with a sensor with a so called bayer pattern, only measures R, G or B values per cell. That means that per RGB pixel, there are two missing values that need to be interpolated. This is also called demosaicking. There are various algorithms do to this, but the key idea, is that interpolated pixel values are related to each other. A relation that might be disrupted by image tampering.

Everything and more on this subject can be found at the university of Dartmouth.

[ash]

Whoa.. this is interesting stuff!

I don't know much about image processing, but you also have to remember that random noise will invariably affect the picture too. So its quite hard to know if variations are caused by deliberate morphing or noise.

Is it possible to take pictures and output it as vector images? I bet that will take a lot of resources and mathematical power to do so The picture would have to be taken by a special camera or something.. but it would still be cool nonetheless!

[xheavenlyx]

There are various algorithms do to this, but the key idea, is that interpolated pixel values are related to each other. A relation that might be disrupted by image tampering.

Ohhh man, this is a really cool idea!!! The interpolated pixels will have a mathematical relation with their immediate neighbor!

The link you have given is invaluable to this discussion!!

And ash, yea noise will effect a picture and must be considered in ALL our assumptions.

As for vector thats a good idea and i have no idea how it would be possible.

[The_Big_K]

Originally Posted by xheavenlyx

Ohhh man, this is a really cool idea!!! The interpolated pixels will have a mathematical relation with their immediate neighbor!

The link you have given is invaluable to this discussion!!

And ash, yea noise will effect a picture and must be considered in ALL our assumptions.

As for vector thats a good idea and i have no idea how it would be possible.

[spam] He's an 'Imaging' engineer! Did you notice that?! [/spam]

[ash]

Oh yea, I just realised that! Thats a cool title

Slightly offtopic, I realised that stating your specialization to others will make for a good topic for discussion. Everyone (that is, non-engineers) knows what mechanical, electrical, electronics, communications, computer or chemical engineer, etc are, because they can visualise a "product" or "device" that relates to the discipline. Examples include a Car, plane, petrol.

But if you say your an instrumentation, control or in this case an Imaging engineer, they will be like.. "WOAH.. that sounds so awesome. So what exactly do you do??" And try to visualize the latest gadget or something.

Okay, I'm jealous. So what

[xheavenlyx]

Oops...

Nice to see a professional here

(Maybe a professor!?)

[hbk]

hi.

if anybdy has any links related to "detection of tampering in images",which can be implemented using MATLAB, please post them here...

any help will be highly apprecited..

thx...