Image Splicing Detection Using Electromagnetism-Like Based Descriptor

Abstract

This study proposes a simple and powerful descriptor called Electromagnetism-like mechanism descriptor (EMag) for image splicing detection. EMag is based on the electrostatic mechanism that represents the image pixels as electrical charges. For a given tampered image, the EMag algorithm divides an image into blocks and then calculates the final attraction-repulsion force between the central pixel of the square image block and its neighbors. The experimental results using an image splicing dataset provided by Digital Video and Multimedia Lab at Columbia University (DVMM) confirm that EMag impressively outperforms the other widely used descriptors for the detection of image splicing. Support vector machine is used as a classifier that distinguishes between the authentic and spliced images. The experimental results presented demonstrate that the achieved improvements are compatible with other splicing detection methods.