A New Algorithm to Exposing Image Forgeries by Detecting Ambient Illumination Consistency

Illumination consistency makes one of important roles in image authenticity, especially in splicing and copy-move rendering. This paper proposes a new method to estimate illuminant direction based on idea of binary sort tree. The algorithm calculates the least error function to structure the binary tree of illuminant direction. And it doesn’t need to expand much time to solve pathosis equation and access accuracy is improved. The experiment shows that, the algorithm can effectively calculate the light source direction of different objects and backgrounds in an image especially in infinite light source, and the detection correct rate reaches 85.4% in general situation , what’s more, 88.1% in infinite light source. Anyway, the algorithm in this paper is a great advancement of blind forensics based on consistency in lighting direction.