Automated image splicing detection from noise estimation in raw images

Splicing is a common image manipulation technique in which a region from a first image is pasted onto a second image to alter its content. In this paper, we use the fact that different images have different noise characteristics, according to the camera and lighting conditions during the image acquisition. The proposed method automatically detects image splicing in raw images by highlighting local noise inconsistencies within a quadtree scan of the image. The image noise is modelized by both Gaussian and Poisson noise components. We demonstrate the efficiency and robustness of our method on several images generated with an automated image splicing.