Detecting digital image forgery in near-infrared image of CCTV

Abstract

The reliability of CCTV digital images is more important than the reliability of many other types of images. However, image editing tools such as Photoshop make this unreliable. CCTV uses two photography modes, the RGB mode and the near-infrared mode. While near-infrared images have different properties, such as a constant level of source light intensity, and a constant direction of the source light, there are no forensic techniques for near-infrared images. In this paper, we propose a forensic technique based on a constant direction of the source light. In order to expose splicing forgery in near-infrared images, we create an ideal near-infrared image model of a plane. We then calculate gradient vectors of the model and objects in images. Depending on the similarity of two vectors, the image is determined forged or not. This forensic technique helps to improve the reliability of near-infrared images.