Research On Image Forensics Of Identifying CG/PR Image Based On Color Quaternion Wavelet Transform

Color Image is now one of the most widely used image carrier, it is widely used in information dissemination, social networking, and other fields. With the computer image generation technology (CG) of the rapid development and popularity, is now using a variety of image generation software to generate an image modeled on real scene has become increasingly easy. As digital imaging applications more widely, and as a means of tampering with a computer-generated image, it has become an important part of digital image forensics research. So the study of how to use computers to quickly and accurately distinguish between computer-generated image and the natural image (PR) has important research value and significance. This paper studies based on color quaternion wavelet transform (CQWT) digital image forensics technology, the main contents are as follows:Firstly, quaternion wavelet transform QWT (Quaternion Wavelet Transform) based on the theory proposed color image forensics scheme QWT is based. Similarities and differences between the analysis QWT discrete wavelet transform DWT (Discrete Wavelet Transform) is compared. Higher order wavelet analysis wavelet statistical features and high-end features to predict the characteristics of the analysis of its application in QWT the details of the office. Simulation results show that the model to achieve improved CG/PR evidence classification accuracy, reducing the amount of computation data and feature dimension to some extent, while the analysis of experimental results obtained each quaternion wavelet sub-band and quaternion argument magnitude performance comparison in image forensics.Secondly, color quaternion wavelet transform CQWT (Color Quaternion Wavelet Transform) based on the theory proposed color image forensics scheme CQWT based. By studying the order of one to four quaternions central moment, the high-order wavelet statistical features and high-end features extended to the wavelet prediction quaternion field of digital image forensics technology to provide technical support in the application of quaternion field. Simulation results show that quaternion forensics features proposed features compared to traditional forensics, both feature dimensions or evidence classification accuracy, have a greater advantage. Robustness tests also show that the proposed scheme has better anti-attack capability, resistant to most anti-forensic attacksAt last, in Markov Markov chain characteristics based on the theory presented based on quaternion Markov statistical characteristic quantities of the color image forensics program. The statistical characteristics of Markov extended to four metadata fields proposed quaternion Markov statistical characteristics for the proposed forensics program provides three-dimensional features. Experimental results show that the CG image forensics, quaternion Markov statistical characteristics have certain advantages.