Exposing Copy-Move Forgery Based on Improved SIFT Descriptor

Abstract

Copy-move forgery is one of the most common forgery methods. When creating a copy-move forgery, it is often necessary to add or remove important features from an image. To carry out such forensic analysis, various copy-move forgery detection (CMFD) methods have been developed in the literatures. However, most of them suffer from their inability to detect the duplicated region, which is subjected to various transformations such as scaling, rotation, flipping and blurring. In recent years, some forensic approaches using the feature-based technique to detect copy-move forgery. Although these methods are proved to have robust performance in this field, when the duplicated region is of uniform texture, this kind of methods normally failed to detect such forgeries due to insufficient or even none key-points located. Furthermore, many feature-based CMFD methods suffer from the problem of insufficient matched key-points while performing on the mirror transformed forgeries. In this paper, a novel key-points distribution strategy is developed for interspersing the key-points evenly throughout an image. The proposed method is able to deal with the problem of insufficient key-points in the almost uniform area. An improved scale invariant features transform (SIFT) descriptor is used to increase the accuracy of CMFD in rotation and mirror reflection situation. Experimental results show the efficacy of this technique in detecting copy-move forgeries.