Image Processing Reference
In-Depth Information
a Forged image (005_F)
b Expected result (005_B)
c Actual result
Fig. 6.11
Detection of blurred image
6.5
Future Work
The current method is developed with the aim to cope with detection of plain CMF.
However, beside simple plain CMF, there are many different variations of forgery
where some additional transformation of the copied region is applied. Transforma-
tion can include some linear change of size (scaling) or orientation (rotation) of
the copied region. Furthermore, transformation can be done as a nonlinear function
where the transformation factor is not the same in all directions. Finally, more than
one transformation can be applied to the copied region.
The problem of detection of transformed copied area requires the definition of
more complicated neighbourhoods. Detection of rotated regions can be solved by
using a circular neighbourhood for every block of image. The learning process
would remain the same as in the case when 1D CA is used. On the other hand,
scaling of the copied region would require using more than one neighbourhood for
each block. Those neighbourhoods should have different size to allow detection of
changes in sizes of the copied region.
Furthermore, transformation of the copied region introduces small changes in
pixels values due to intensity interpolation effects. Those changes do not impact
the global texture of the copied region, but they affect the local thresholding process
when neighbourhood binary values are defined. Coping with this problem is possible
by application of more advanced thresholding methods for representation of images
in a binary way.
6.6
Conclusion
Digital image forgery has become a huge problem due to the simplicity of pro-
cessing of digital images. However, techniques for detection of forged images were
introduced in parallel with the development of forgery attacks. Detection methods
have some kinds of advantages and disadvantages, but there is still no unique method
for detection of all kinds of forgeries. One very common forgery type is copy-move
forgery (CMF) where part of the image is selected, copied and moved to a new
Search WWH ::
Custom Search