Research On Splicing Detection For Color Images Based On Quaternion Features

Along with the advance of image processing technology, there are more and more tampering images in our life. Splicing is one of main means of tempering, which is used to modify the content of the image to confuse and deceive people. Aiming at this problem, there have been a lot of schemes for splicing detection. However, most schemes are based on graylevel image or use one of color channels for splicing detection. It results that features of these schemes are short of color images' spatial structure, the channel correlation and color information.In this paper, the quaternion matrix is used to represent a color image. Schemes based on quaternion transform domain for color image splicing detection are proposed.The main work is as follows:1) In order to retain the color information of color images, color image splicing detection based on QDCT domain is proposed. In the proposed scheme, the color image is first divided into non-overlapping blocks. Each block is applied with QDCT. Then, Markov features are extracted from QDCT coefficients. At last, the features are used for splicing detection. In this scheme, the magnitude and phases are used to represent the luminance, spatial and texture information. On this basis, three ways for threshold process are proposed. The experiments show that the accuracy for splicing detection arrives at 99.16% and 97.52% in CASIA TIDE v1.0 and v2.0, respectively, which is better than most of the existing schemes.2) To solve the loss of block correlation and global information of QDCT, color image splicing detection based on CQWT and QDCT is proposed. In this scheme,quaternion Markov based on CQWT is proposed. Quaternion Markov features are extracted from phased, which is combined with features based on QDCT. The experiments show that the accuracy for splicing detection arrives at 99.47% and 98.13% in CASIA TIDE v1.0 and v2.0, respectively, which proves the effectiveness of the combined features.