IMAGE FORGERY DETECTION USING SUPPORT VECTOR MACHINES: DEVELOPMENT AND EVALUATION

Abstract The revolution in digital imaging technologies has changed our lives. In comparison to textual information, information through an image is more direct and instant. With the help of sophisticated software, a real-like image can be generated and a real photographic image can be manipulated. Hence it is possible that some image forgers can maliciously tamper the digital image to distort the truth. To restore trust in digital images, the investigation of forgery in images is essential. When the content of a digital image is modified or the whole image is created using a computer graphics software for wrong motives, such an image is commonly referred to as a forged image, and the process of creating such forged image is known as digital image forgery. In this dissertation, a systematic and complete solution is provided to authenticate a digital image. For this, two commonly performed image forgery types viz. copy-move forgery and image splicing forgery have been investigated in this thesis. Further, it is also required to develop are liable method to detect image forgery in photographic images. Various concepts such as Tetrolet transform, Gaussian-Hermite moments, Neuro-fuzzy classifier, etc. are explored and used in the proposed approaches. Experimental validation of all the proposed methods has been performed on various available image datasets and the results are compared to the existing technique on the behalf of certain performance parameters including Compression, Robustness, Accuracy and F1 score.