Resampling Tamper Detection Based on JPEG Double Compression

摘要

JPEG compression and resampling operations appear frequently in the forgery image, so the traces of these two operations are used to detect the tamper of images. As JPEG compression and resampling operations affect the linear correlation in coefficients, so the traces of JPEG compression and the linear correlation are introduced when images are resampled and saved in JPEG format. Hence, the moment of DCT coefficient histogram are taken as the factor of JPEG compression, and due to the singular value decomposition can measure linear correlation between pixels well so that the mean value and variance of the singular values are extracted. At last, two types of features are extracted and trained by SVM. Experimental results show that proposed method has a good detection rate for resampling compression images, especially for the rotation operation and the scale factor which greater than 1.