Forensic Detection of Median Filtering by Hough Transform of Digital Image

Abstract

In a distribution of the digital image, there is a serious problem, such as illegal apportionment of the altered image by a forger. To solve this kind of problem, a novel median filtering (MF) detection algorithm is proposed, in this paper, using feature vectors extracted from the peaks of degree and distance parameters in the Hough transform of the image. In the proposed algorithm, the coordinates of the peaks are the feature vectors. The proposed MF detection scheme compares to the MFR (Median Filter Residual) and the VNLP (Variation of Neighboring Line Pairs) schemes that have the same 10-D feature vectors. The defined 10-D feature vector is trained in SVM (Support Vector Machine) classifier for the MF detection of the forged images. The performance of the proposed MF detection is measured at Unaltered, Averaging filtering (3 × 3), JPEG (QF = 90), Downscaling (0.9) and Gaussian filtering (3 × 3) images respectively. Subsequently, in experimental items; AUC (Area Under the Curve) by the sensitivity and 1 - specificity, $P_{e} (a$ minimal average decision error), and $P_{TP}$ at $P_{FP} = 0.01$ are performed. Thus, it is confirmed that the grade evaluation of the proposed algorithm is ‘Excellent (A).’.