A hybrid technique to discriminate Natural Images, Computer Generated Graphics Images, Spliced, Copy Move tampered images and Authentic images by using features and ELM classifier

The paper presents a novel technique in the field of image forensics. The proposed algorithm is able to differentiate between computer generated images (CGIs), natural images (NIs), spliced images, copy moved tampered images and authentic images. The algorithm is also able to localize the tampered region in spliced and copy moved tampered images. At present, most of the detection techniques are able to encounter only one type of tampering at a time but the proposed algorithm can encounter all the three types of tampering simultaneously. The algorithm has two phases, in the first phase the algorithm has used laplacian of gaussian and autocorrelation to differentiate CGIs from NIs. In the second phase, the algorithm has used estimation of color filter array pattern and sensor noise to detect spliced region. The Grey level co-occurrence matrix (GLCM) has used to detect copy moved tampered area. Additionally, the algorithm can detect such tampering in presence of Gaussian blur, JPEG compression and scaling. The algorithm is also able to localize rotated and scaled portions of copy move tampered area. At present, none of the research papers encounter all these three problems simultaneously. The efficacy of the proposed algorithm has been verified through extensive simulation.