Locality Preserving Projection Based Multiple Copy-Paste Forgery Detection

Abstract

Digital images are the dominant media of information in the digital world and used to convey the desired information. It has been noticed that manipulated images are put to wrong use to create a false perception about the image or story in support. Copy-paste forgery is one of the most exploited image manipulation approaches. In recent years the research community has proposed many methods for the detection of such forgery. In this work, the copy-paste forgery has been detected and localized by the proposed scheme using locality preserving projection (LPP). As the processing level increases, changes in pixel intensity and statistical changes in image pixels increase, which reduces the accuracy of localization. Similarity preserving property of LPP is used for developing a multifaceted method for copy-paste forgery detection for images attacked with various post-processing.