The Forensics for Detecting Manipulation on Part of Text

The adventure of sophisticated photo editing software has made it increasingly easier to manipulate text in digital images. Visual inspection cannot definitively distinguish the forgeries from authentic photographs. In the imaging process, as defocusing, atmospheric turbulence, diffraction and the defects of image device, digital image can´t reproduce the details of texture perfectly, while the text image that obtained by photo-editing software will do better, and will show no difference with the authentic image in the appearance. In this paper, we describe a new forensics technique that focuses on manipulation on part of text. By decomposition of the text and to extract the characteristics of edge points in text image, we use support vector machine (SVM) to training classification model which is used for the identification of the authenticity of text message in images. Our test result indicates that this method can detect the manipulation on part of text, and the general detection rate is promised.