Digital Image Forgery detection using color Illumination and Decision Tree Classification

As an image can convey more information than words, many people rely on them for communication. To establish the authenticity that the image is not a composite or spliced image, this paper proposes a new technique that detects the image forgery based on illuminant color estimation. The proposed system estimates illuminant color of the image and creates illuminant maps of human faces in images using Gray world approach and IIC space. It then extracts the texture features of the faces by using an integrated technique which incorporates Gabor local binary pattern for which Histogram of Orientation are derived by HOG Edge algorithm. SASI descriptor is used to calculate the distance in statistical information of the paired faces. The proposed system needs human interaction only to select the human faces on images. As the existing forgery detection methods rely on human experts to take final decision on tampering, this paper includes a classification using decision tree classifier which works by machine learning and eliminates the need for human expert to take final tampering decision. The results of decision tree classifier are potential to indicate the authenticity of images.