Detecting digital image forgeries through weighted local entropy

In this paper we propose a new method that could reveal some traces that were left in the tampered digital images without using watermarking technique. Our method is based on detecting the presence of feather operation, which is particularly useful to create a smooth transition between the selected region and its surroundings. The similarity of pixel neighborhood is used as weight coefficient to modulate local entropy value. The forged region can be determined by the value of weighted local entropy precisely. Results of experiments on tampered images will be given.