Accurate And Efficient Copy-Move Forgery Detection

Abstract

Today, due to the development of image processing applications like GIMP and Photoshop, forging of digital images with the absence of any trace is not at all difficult. Therefore crucial validation of content in a digital image is essential. Among the many prevailing forgery techniques, the copy-move forgery is recurrently occurring and it is easily performed. Detection of copy-move forgery is difficult as the forged region show similar properties to the source image. This paper analyzes the issue of copy-move forgeries and proposes an effective and reliable technique for its identification. Basically for the detection of copy-move forgery, two main approaches are applicable - block based and keypoint based techniques. Block-based technique employs division of image matrix into blocks that overlap and features are extracted. In keypoint based approach features of each keypoint is extracted for further matching. The aim is to implement two algorithms for detecting copy-move forgery - one using block based Discrete Cosine Transform algorithm and the other using the keypoint-based Scale Invariant Feature Transform. Efficiency of both algorithms are measured and their comparison is done using the GRIP and CoMoFoD databases. It was found that the Scale Invariant Feature Transform has more accuracy and precision compared to Discrete Cosine Transform method.