Exposing Digital Image Forgeries by Illumination Color Classification

In this paper, we tend to analyze one amongst the foremost common styles of photographic manipulation,called image composition or splice. We tend to propose a forgery detection methodology that exploits refinedinconsistencies within the color of the illumination of pictures. Our approach is machine-learning primarily basedand needs borderline user interaction. The technique is applicable to pictures containing 2 or a lot of folks and needsno professional interaction for the meddling call. To attain this, we tend to incorporate info from physicsandstatistical-based fuel estimators on image regions of comparable material. From these fuel estimates, we tend toextract texture- and edge-based options that square measure then provided to a machine-learning approach forautomatic decision-making. The classification performance victimization associate degree SVM meta-fusion classifieris promising.