Gaussian Forensic Detection using Blur Quantity of Forgery Image

For a design of the Gaussian forensic detection (GFD) in the altered digital images, this paper presents a feature vector that is defined as a blurring quantity by the window size of the Gaussian filtering. The window size is prepared ten types, and their blur quantity is computed by the Gaussian filtering, respectively. In the proposed scheme of the GFD, the defined 10-dim. feature vector of the image is trained in a SVM (Support Vector Machine) classifier for the Gaussian forensic detection of the forged images. In the experiment, the measured area under the curves (AUC) are above 0.9 from a classification point of view between the altered and Gaussian filtered image. Thus, the grade evaluation of the proposed method is rated as "Excellent (A)."