Content area
Full Text
ABSTRACT
Image splicing is to compose a new image from two or more images, and it is widely used for image forgery. Image splicing detection is a key problem in image forensics. However there are very few solutions to this problem. These solutions have a low detection accuracy, and most of them are not able to expose the spliced region. This paper proposes a novel image splicing detection method based on the assumption that images produced from different cameras use different demosaicking algorithms. The proposed method can effectively detect the spliced images, and expose their corresponding sliced region. Experiments show the proposed method has a high detection accuracy of 91.1% on Columbia Image Splicing Detection Evaluation Dataset. This paper also shows the procedure to detect the exposed region on a spliced image.
Index Terms- image forensics, splicing detection, demosaicking inconsistency
(ProQuest: ... denotes formulae omitted.)
1. INTRODUCTION
In modern world, digital cameras and smart phones produce an enormous number of digital images. These images pro- vide convincing evidence in all aspects. Meanwhile with the ease of the image manipulation tools, such as Photoshop and Gimp, digital images are more easily to be modified. Thus the phrase 'Seeing is believing' may no longer hold.
To make sure that an image is authentic, researchers pro- pose two directions : active watermarking [1] and passive im- age forensics [2]. Watermarking actively requires embedding some message into images when they are generated or before they are distributed. However, this requires additional hard- ware, thus most imaging devices do not carry this function. Image forensics makes use of the characteristics of the imag- ing device and the properties of digital images to test whether images in question are authentic, in particular whether they have been manipulated. Among all the image manipulating operations, composing a new image from two or more im- ages, referred to as 'image splicing', is a common operation.
Image forensics is still in its infancy. Although splicing detection is a key problem, not a lot of work has been done. Hsu and Chang [4] check the consistency of camera charac- teristics among different areas in an image, but they only re- ceived 70% precision. Chen et al. [5] extract features from the sharp transitions introduced...