When an input image, the system will first conduct initial face model fitting, creating a low-frequency albedo map estimation. Then, they run a texture analysis, with information from the face ...