Using the local information of image to identify the source camera

In this paper we introduce a new method for source identification in digital image forensics. The proposed method uses local information of the inherent pattern of the camera, as a signature of the camera for source identification. Here the sensor pattern noise is used as the unique identification property of the camera. However due to content dependency of the denoising algorithms that are used to extract the noise pattern, the different regions of the image do not have the same information about the camera signature. Hence in our algorithm, at first the best regions of the image according to their local information are selected to extract the noise pattern. This step is done by fuzzy-based classification on the overlapped blocks of the image. In the next step the noise pattern of these regions are extracted and then, we evaluate the correlation between the image pattern and camera pattern. Finally the source camera is determined according its correlation. The experimental results compared to similar works show an increase in the detection rate of source identification, while computational complexity is reduced; this affirms the efficiency and performance of the proposed theory.