A method of identifying digital images with geometric distortion

Abstract

In this digital era, anyone can easily tamper with an original image using various editing tools. One problem is that tampered images are being used for crimes such as the violation of portrait rights and the forgery of official documents. So far, various studies have been investigating a method for identification of the tampered images. However, these studies cannot detect the tampered region when geometric transforms are applied to the original image, because the geometric transform involves various interpolation methods. In this paper, we propose a new method for detecting the tampered region. First, the interpolation is applied to the image. If the image is tampered with by geometric transformation, then this re-interpolation has a role of emphasising the tampered region. Secondly, we create a detection map in the frequency domain using the re-interpolated image, which is useful to intuitively identify the tampered region. We tested our proposal on two kinds of image. One is entirely changed by geometric transforms, and the other is partially changed by geometric transforms. Our experimental result shows that we can easily catch not only the tampered region but also the distortion direction.