A study on pre-processing methods for copy-move forgery detection based on SIFT

Copy-move forgery detection is one of the most popular type of forgery in multimedia forensics. The attacks copy and paste desired area in the same images. In this paper, we aim to reduce the false matches of detected area of forgery by using the pre-processing methods, such as histogram equalization, smoothing filter with median and Gaussian filter and sharpening filter, combined with the SIFT algorithm. We presented the results of each of the pre-processing methods for total SIFT keypoints detected, total number of matches and total number of false matches. From 30 tampered images that fell into 3 groups, the SIFT method could be combined with pre-processing methods to reduce false matches and the best pre-processing method that gave the lowest false matches results was dependable on the types of images in the database.