A Study On Quaternion For Color Digital Image Forensics Technology

In recent years, digital multimedia such as digital images, video and audio has been widely used, at the same time, multimedia content is easily tampered by using multimedia editing software. Thus how to ensure the safety of digital multimedia content has become the focus of many scholars. Digital multimedia forensics technology, which is mainly used for the identification of digital content authenticity and integrity and for the copyright protection, is a branch of information security. Digital multimedia forensics technology mainly includes the active and passive forensics. The typical active forensics includes digital watermark and robust image hash, while passive forensics is referred to as blind forensics, which does not depend on any original digital media content. In this dissertation, some basic theory and method were investigated including quaternions, quaternion Fourier transform, quaternion Zernike moments, uniform log-polar mapping, invariant feature of key points, copy-move tampering forge detection model and KD tree, and several novel digital image forensics methods were proposed for copyright protection, image content authentication, tampering detection, localization and copy-move forge detection. The main contributions are summarized as follows:First, a blind color image watermarking scheme based on quaternion discrete Fourier transform (QDFT) and on an improved uniform log-polar mapping (IULPM) for copyright protection is proposed. Existing robust to resist geometric attacks of digital watermarking technology is mainly applied in gray image, and a few methods are conducted in color image. Similar to Fourier transform and log-polar coordinates transform, both Quaternion Fourier transform and log-polar coordinates transform had the same property as that may simplify the rotation and scaling of image in the space domain to shift in the frequency domain, thus they were applied to the robust color images watermarking in this dissertation. Secondly, a high accuracy of watermarking synchronization method based on IULPM was proposed to resist the fractional angles of image rotation. The proposed watermarking scheme embedded dual watermarks:one was a meaningful binary watermark and the other was a bipolar watermark. The former was embedded into the real part of mid-frequency QDFT coefficients using quantization index modulation (QIM). The latter was used for watermark resynchronization after the watermarked image has been attacked, making the scheme resistant to geometric attacks. At the same time, the watermark embedding employed the entire image instead of block manner. Experimental results demonstrated that the proposed scheme achieved better robustness against common signal operations and geometric attacks compared to some existing schemes.Second, a novel robust image hashing scheme based on the combination of the quaternion discrete Fourier transform (QDFT) with the log-polar transform is proposed for image authentication. Existing color image hash methods transform the color image into grayscale one or deal with the three channels of color image simultaneously, which could lead to the loss of chromatic information or enlarge the hash length. QDFT offers a sound way to jointly deal with the three channels of color image so as to overcome above issues. The key features of the proposed method relied on (i) the computation of a secondary image using a log-polar transform; and (ii) the extraction from this image of low frequency QDFT coefficients’ magnitude. The final image hash was generated according to the correlation of these scrambled magnitude coefficients. Experiments were conducted in order to analyze and identify the most appropriate parameter values of the proposed method and also to compare its performance to some reference methods in terms of receiver operating characteristics curves. The results showed that the proposed scheme offered a good sensitivity to image content alterations and was robust to the common content-preserving operations, and especially to large angle rotation operations.Third, a novel robust image hashing scheme based on quaternion Zernike moments (QZMs) and the scale invariant feature transform (SIFT) is proposed for image tampering detection and localization. The proposed method not only can overcome the problem of a long hash length and the loss of chromatic information, but also can locate the tampered region and detect the nature of the modification, including object insertion, removal, replacement, copy-move, and cut-to-paste operations. QZMs considered as global features were used for image authentication while SIFT key-point features provided image forgery localization and classification. The proposed method compared with several recent and efficient methods. These experiments showed that the proposed scheme provided a short hash length that was robust to most common image content-preserving manipulations and large angle rotations, and allowed us to correctly locating forged image regions as well as detecting types of forgery image.Fourth, a copy-scale-move (CSM) forgery detection method based on a pyramid model is proposed. In block based division approaches, existing copy-move forgery detection method is not robust to the scaling operation of region replication. In order to detect image forgery of copy-scale-move (CSM), a test image was first rescaled into down sampling sub-images with different scale factors, which were used to build the pyramid model according to non-uniform sampling. Then, each sub-image in the pyramid model was divided into overlapping sub-block. The Zernike moment features of all sub-block were extracted. Finally, the block features of each sub-image and those of the original image were used to build a KD-tree to search all approximate matching pairs, from which the forged trace "shift vector" can be found so as to determine the tampering region. The experimental results showed that, compared with the existing image tampering detection methods, the proposed method provided better detection results for the precision and recall.This paper provided several effective methods for digital image forensics. However, with the development of sophisticated image editing processing software and anti-forensics technology, the scenario of image tampering detection will be more complex and difficult. Thus it is a challenging task to further perfect and improve the digital image forensics technology for a long time.