Exposing Fake Images Using Graph Theory

Main Article Content

Dr. S. Pariselvam, D. Arivoli, Deepak Vengatesan

Abstract

Detection of fake images using graph data structure which have a tendency to decide the rhetorical Similarity in the image, which captures key rhetorical relationships within the regions in the image. The image is first broke-down into a small blocks called patches to compare the similarity among the pixels using iterative graph data structure and then create cluster of patches with reference to their similarity. If the image is being spliced the cluster will contain multiple segments of traces that are been record from the image structure called FruchtermanReingold structures which shows the overlapping image segments in the image. This FrunchtermanReingold structure consist of vertices and edges by using spectral clustering algorithm we can differenciate the original region and the duplicate region of an image which can be then mapped back to the image using igraph algorithm. This methodology would troune existing art pictures that weren’t able to capture the spliced traces of a picture. As a result, identification of a spliced regions by the multiple commnunities presented in a picture and constraint is performed in a picture and constraint is performed by partitioning these spliced image traces.

Article Details

Section
Articles