Improved 3D lighting environment estimation for image forgery detection

3D lighting environment is an important clue in an image that can be used for image forgery detection. Existing forensic methods exploring lighting environment consistency are based on many assumptions, among which convexity and constant reflectance of the surface are two critical ones. In this paper, we propose an improved 3D lighting environment estimation method based on a more general surface reflection model. We relax the two assumptions by incorporating the local geometry and texture information into our position dependent reflection model. The proposed model is more realistic for objects like human faces which are non-convex and textured. Experiments show that the proposed method can achieve improved lighting environment estimation accuracy compared to the previous method and has better forgery detection efficacy.