| Along with the popularity of image processing software,it is easier for people to edit an image.There are many forgery methods and copy-move forgery is one of main means of tempering.Forgers copy some regions from an image and past them to the same image to cover or enhance something in the image.So far several copy-move forgery detection methods have been proposed.However,most of the methods covert color images into gray ones.It results that color images' spatial structure,the channel correlation and color information can not be full used,Aiming at this problem,we proposed several copy-move forgery detection methods based on quaternion transform.The main work is as follows:1)Aimed at the problem that the color information of images is ignored,in this paper we propose a forgery detection scheme based on quaternion discrete cosine transform(QDCT).Firstly the image is divided into overlapping blocks.Then the QDCT is performed to these blocks.Then we extract the feature of these blocks by zigzag scaning.Then the feature is lexicographical ordered.At last,we find the block pairs whose Euclidean distance is less than the preset threshold in the consecutive rows in the matrix and distinguish these blocks as forged region.Experiments show that the missing-false alarm rate in the method based on QDCT is lower than the existing methods.2)Aimed at the problem that the color information of images is ignored,a copy-move forgery detection scheme based on quaternion principal component analysis(QPCA)is proposed.The proposed scheme is block-based and firstly it divides images into overlapping blocks and performs QPCA of all the blocks to extract QPCA features.Then the features are lexicographical ordered to obtain shift vector and its frequency.Finally we compare the shift vector frequency to the threshold to locate the forged region.Experiments show that the missing-false alarm rate in the proposed method is lower than the existing methods and has a better accuracy. |
PDF Full Text Request