Image Forgery Detection Based on Semantics

Abstract

The development of powerful image editing software has made it easy to create visually convincing digital image forgeries. Recently some research works based on general low-level visual features in the field of digital image forensics have been conducted to address this problem. However, there has been little work by analyzing the high-level semantic content of the image. This paper discusses the forgery detection problem of digital image from the point of high-level semantic, and proposes the framework of image forgery detection based on semantics. The framework consists of three components including image recognition, semantic logic reasoning engine and generation of semantic rule. A new fuzzy logic semantic reasoning model for image forgery detection is proposed in this paper. Our reasoning model based on fuzzy logic can give the deterministic and uncertainty reasoning results. Based on the proposed framework of image forgery detection, the prototype system of image forgery detection based on semantics is completed using MATLAB, MySQL database software and ontology software protege and has the ability to detect abnormal splicing from the foreground and background, avoiding detection using the traditional methods based on lowlevel image features. Experimental results show that the framework and logic semantic reasoning model for image forgery detection are effective.