How math can tell you if a picture is doctored
This is a picture of a cat, obviously, but do you notice something strange about it? Sure looks like a cat, but let's have a closer look - zoom in!
This is a picture of a cat, obviously, but do you notice something strange about it? Sure looks like a cat, but let’s have a closer look – zoom in!
I mean, really zooooom in.
Once you zoom enough any picture, you’ll end up with a regular rectangular mesh of cells called pixels, each pixel containing a colour value. The image below shows how each pixel looks like when encoded by colour.
Now we’re talking business! Ann Dooms’ business that is. Dooms is a mathematician, but also a forensic specialist. She uses her mathematical skills to spot doctored photographs and videos by analyzing the mathematical patterns each of these digital mediums share or, on the contrary, lack. Algorithms can be used to find “patterns, irregularities, additions, and changes,” she says. Let’s take these two side-by-side photos below, can you spot the difference? (hint: don’t try too hard).
Again, in colour code.
As you can see, some numbers are off which means the two photos are distinct.
Using this sort of simple method, one can easily determine whether a photo or video was doctored, greatly aiding in criminal investigations such as cases of fraud. Fake paintings are also spotted this way. And that’s not all – feel welcome to Dooms’ TED Talk below to find out much more about her inspiring work.
