A blind forgery detection scheme using image compatibility metrics

In order to check the integrity of a digital photograph, a blind forgery detection scheme for forensics is presented in this paper. The proposed scheme is exposing image fakes by using two different terms of compatibility metrics, such as inconsistencies of edge blur width and lighting direction, respectively. The inconsistency metric of an edge blur is constructed based on the deviation from its linear fitting by measuring the blurriness of each pixel along a doubtable edge. Then such metric is used as an evidence for identifying discontinuity of edge in image slicing manipulation. When creating a digital composition of images, it is often difficult to exactly compose the lighting effects due to directional lighting. The lighting inconsistencies can be an useful tool for revealing traces of image composition. Therefore, the direction of the lighting source can be estimated for different objects/people in an image. In fact, the inconsistency of lighting direction has a close relationship with image illumination, which is then used as an evidence of image composition. Experimental results of the proposed scheme on natural images are presented for comparing with other related methods.