Ever since the popularisation of digital imaging, the
modification of images has been made more and more
accessible, even to non-professionals. In this context,
it becomes important to be able to detect such alterations, especially in cases where images and videos are
involved in legal proceedings and could be used as evidence. In this paper, we propose a novel approach for
detecting image compositing in video images based on a
realistic CCD sensor noise model developped by Healy
and Kondepudy [4].
There are relatively few papers related to this field.
Popescu [10] detects re-sampling and rotations with an
EM algorithm. Fridrich, Soukal and Lukas [7] propose
a method for detecting cases where part of an image
is hidden by copying another part of the image over
it. They do this by comparing square blocks of pixels
within the same image. Popescu [10] improves on this
method with a statiscal analysis by comparing properties of blocks. Farid and Lyu [2, 11] propose a method
based on wavelet decomposition of images for differenciating natural and sythesized images. Johnson and