Using FRFT And NMF Image Copy - Paste Tamper Detection Algorithm

The rapid popularization and application of intelligent terminal devices, such as mobile phones, tablet PC, digital TV, greatly speed up the speed of digital videos, images, audios, etc, which make people’s life colorful. With the widely application of the increasingly powerful, easy-to-use digital image editing software, people who don’t need special professional technology can lifelike tamper with digital information. The events of malicious tamper with image occur frequently, which have a bad effect on society and let people lose the trust of authenticity of the image. For protecting copyright and preventing image forgery or alteration of document with illegal intentions, the tasks of digital image forgery detection has became more and more important research subject.Image forgery detection belongs to the category of digital image forensics technology. Image forensics can be divided into two classes:classes--active authentication and passive-blind authentication. Active authentication is the method that prior to add digital signature and watermarking in the digital image. In practice, most of the digital images are not add any watermarking or signature information. Now, passive-blind authentication has become the mainstream of image forgery detection research. In this paper, we concentrate on detection of copy-move forgery detection in the image, where a part of image itself is copied and pasted on another part of the same image. Copy-move forgery is one of the most simple and common used image forging methods.This paper proposes two methods of image copy-move forgery detection. One of them is a new algorithm for copy-move forgery detection which based on fractional Fourier transform (FRFT). Firstly, the1-level DWT is used for the image to reduce the image dimension and computational complexity. The low-frequency of the sub-band after transformation is divided into the same size, overlapped sub-block. Then FRFT is applied for each block to calculate the vector of the coefficients. Finally, the difference of feature vectors is evaluated between two adjacent vectors after lexicographic and employed to locate the duplicated regions which have the same feature vectors. The process of image copy-move can be seen a process of nonlinear. As a method of adaptive time-frequency analysis, FRFT has a lot of advantages in analyzing and processing nonlinear signal. Experimental results show that the proposed method is effective in detection of the image copy-move forgery regions. Another simple method in this paper is applied Non-Negative Matrix Factorization (NMF) which is based on exact match[14].The algorithm converts each sub-block after overlapped into a row vector and then form a new matrix. NMF is applied for the new matrix to obtain a base matrix and a coefficient matrix. Lexicographical sorting is used for the base matrix which is regarded as feature vector. At last, the difference of adjacent vectors is computed and forged areas which have the same feature vectors are located. The algorithm is easy to understand and proved effective. The running time of this algorithm is reduced.