Perceptual Hash Based Detection Algorithm for Image Copy-Paste Tampering

Abstract

One of the most common modes of image attacking is copy-paste tampering. To overcome the disadvantages of slow speed and fragility in most existing regional tamper detection algorithms, a detection algorithm based on Perceptual Hash Algorithm (PHA) is proposed in this paper. In the aspect of feature extraction, the algorithm firstly carries out Discrete Wavelet Transform (DWT) on the test image and then selects the approximate sub-band to overlapping. Finally, the PHA is used to extract sub-block features and generate the feature matrix. In the patch matching phase, the first step is to find out the similar patches whose statistical displacement is larger than the threshold, and then obtains the main shift vector which satisfies the frequency of occurrence. Finally, the feature matching is performed according to the Hamming distance of similar patches. The experimental results demonstrated that the proposed algorithm can accurately detect the tampered region while ensuring fast computing speed and is robust to some common post-processing operations such as JPEG compression, Gaussian noise and Gaussian blur.