With the use of powerful image modifying softwares, image authenticity is a big question for image forensics. One can no longer believe what they see. When a section of image is copied, geometrically transformed and pasted at different spot onto the same image with the intention of concealing or hiding some important information, it is copy move forgery. In the past few years several techniques for copy-move forgery detection have been proposed. In this paper Discrete Wavelet Transform (DWT) have been used with Scale Invariant Feature transform (SIFT) for copy move image forgery detection. SIFT keypoint descriptors are extracted from the low frequency subband of the discrete wavelet transformed image. The extracted keypoints are grouped into clusters using either of the linkage methods (median, centroid or ward) and are matched to detect the forgery. Different wavelet bases with SIFT have been compared using True Positive Rate (TPR) and False Positive Rate (FPR) as the performance evaluation parameter along with the computation time on a wide range of forged and original image database.

SIFT Copy-move forgery invariant keypoint