fake images strengthened
of photographs faster and more accurately than current methods.
|Allan Swart / 123rf|
Researchers at India’s Jaypee University of Information Technology have
developed a new way to detect ‘copy-move forgery’ in photographs that is more
successful and faster than currently available methods.Copy-move forgery is difficult to detect because
it involves copying an object or an area in an image and using it in another
part of the same image. This is typically done to cover something up or to add
something that was not previously there. For example, an image of three missiles
can be forged to look like an image containing five, where two of the missiles
are copied and added elsewhere in the picture. Forgers can also use this
technique to hide critical data in a crime scene.
Copy-move forgery is difficult to detect because
the altered parts of the photograph contain a similar palette, texture and
distortions as the original image.
Currently available approaches to detect this
type of forgery are slow, as they involve a large number of computational
calculations. They can also give false positive results.
The researchers sped up the process by
converting an image into its binary form, such that each pixel in the image is
either black or white, with black representing parts of objects in the image
and white representing the background.
The team developed an algorithm that then
translates each black pixel in the image into a position on a histogram. Copied
parts of an image will have the same ‘profile’ in the histogram as an original
object in the image. This appears in the shape of a repeated ‘valley’ in the
histogram. Other objects in the image will also appear as valleys, but they
will not match the other valleys.
The team successfully tested their method on
more than 20 forged images of various sizes and found that there is “tremendous
improvement” in the computation time, especially for larger images,
compared to other methods.
The technique is limited to images with distinct
contrast between copied objects and the image background, according to the
study published in the Pertanika Journal of Science & Technology.