科研之友
微信
新浪微博
Facebook
分享链接
摘要
In modern society, as the important medium of information transfer, digital image plays a more and more important role in our daily life. With the modern science and technology revolution, a social phenomenon that people without any professional technique can easily forge and process digital images become commonplace. The image composite forgery, also called copy-move forgery, is the most popular image forged operation. Mostly existing methods are inept for the detection of the composite forgery image underwent geometric distortions. This paper presents a robust and efficient analytical Fourier-Mellin transform (AFMT)-based method. The focus of AFMT method is to construct the scaling and rotation invariant and extract its invariances for the detection of composite forgery. First, the general AFMT expression is given. The radial complex exponential kernel of AFMT is discussed to get the orthogonal feature. Then, the invariant to scaling and rotation is presented to construct the image geometric moment invariants. To extract the scaling and rotation invariance of each pixel of detecting image, a disk template is applied for sliding on the detected image and calculating geometric invariant features. After extracting geometric features, useful geometric features are further filtered from image background information. Then, correlational features of pixels are sorted by lexicographic sorting. Pearson correlation coefficient is applied for identifying the similar continuous regions and locating their positions. Finally, the detected suspicious composite regions are marked. Extensive experiments have been performed to show that the presented AFMT method can detect the composite region in the forgery image precisely. It is also proven that it is more robust and efficient than other existing relevant methods.
- 出版日期2016-4
- 单位广东外语外贸大学
