Publication

Abstract : Image splicing is the art of combining portions of two or more images as if the final image looks like an original single photograph. With the advent of high-quality photograph-editing tools, image splicing is so easy and it is becoming increasingly difficult to detect tampering with the naked eye. This limits the value of accepting photographs as proof in a judicial system. So, developing accurate tools and techniques for splice detection and its localization has enormous potential, especially in digital forensics. In this research work, we develop a deep learning system that makes use of depth information to improve the accuracy of splice detection and localization. We show that the developed end-to-end system surpasses the accuracy offered by state-of-the-art networks Xavier-CNN and SRM-CNN in detecting and localizing image splicing.