Forgery Detection in Digital Images by Illumination Color Classification Using Adaboost Classifier

Photographs have been used to document space-time events and they have often served as evidence in courts. Powerful digital image editing software makes image modifications straight forward and this undermines the trust in photographs. Forgery detection method is proposed which exploits subtle inconsistencies in the color of the illumination of images. It is machine-learning based approach and it requires minimal user interaction. Physics and statistical based illuminant estimators include the generalized gray world estimates and inverse intensity chromaticity estimates on image regions of similar material are incorporated. From these illuminant estimates, texture- and edge-based features are extracted by SASI algorithm and HOG edge algorithm which are then provided to a machine-learning approach for automatic decision-making. The extracted features are then paired using the same descriptors which are then classified by Adaboost classifier.