Search

This paper describes a steganographic method of embedding a binary digital watermark into the mid-frequency coefficients of discrete cosine transform of a digital image. One of the mid-frequency coefficients of the two-color components of a color image is used to embed the digital watermark: one color component is used to embed a secret coefficient which reflects changes in the image and helps to improve the quality of the extracted digital watermark after applying noise to the stego. The second color component of the container is used for embedding of the digital watermark. The proposed method involves the automatic selection of color components of the container for embedding the digital watermark and its pre-processing to increase the distance between its elements. The article contains the main steps of embedding and extracting of a digital watermark, as well as the results of computational experiments aimed at evaluating the efficiency of the steganographic method, and comparing the results with modern analogues. The efficiency of the steganographic method is evaluated by the indicator of visual quality PSNR of stego and various indicators of message extraction. Resistance to attacks is evaluated by indicators of extraction of information from the modified stego. The paper analyzes the developed method when embedding some secret coefficient equal to 20 and the digital watermark into the coefficient (5.5) of the discrete cosine transform of each 8»8 block of digital image is carried out. The results of computational experiments showed the high visual quality of the received stegan messages (PSNR is 50-58 dB) and the resistance of this method to noise, namely: multiplicative, Poisson, Gaussian and "salt and pepper". In the case of Poisson noise, the digital watermark is extracted almost without loss, the quality of extraction of the digital watermark was not lower than 0.97. [ABSTRACT FROM AUTHOR]

In this paper we study the effect of perturbing influences aimed at the spatial domain of digital images on the change of color triplets in the matrix of unique colors of a digital image. In the course of computational experiments established that the modification of only blue color matrix of the image, the quantity of Blue-triads (which indicate the blue color component of the image) in the matrix of unique colors of the digital image significantly exceeds the quantity of Red- and Green-triads that will allow for the examination of images clearly establish the fact of modifying the blue color component. The identified feature is typical for most containers in a lossy format which is associated with the digital image's formation technologies and compression algorithms therefore despite the low sensitivity of the human eye to blue color it is not recommended to use the blue color component for spatial transformations of the image, in particular, for steganographic transformations. During the experiments the most suitable for spatial transformations green and red color components revealed, since their modifications lead to the appearance of a smaller quantity of Green- and Red-triads (which indicate the green and red color components of the image, respectively) and to obtain a comparable quantity of all types of triads of triplets in a matrix of unique colors which will complicate the correct examination of a digital image from the point of view of unambiguously identifying a modified color component. The obtained results are substantiated from the standpoint of color perception by the human eye, which underlies the formation of colors by digital cameras, and the image compression features of the JPEG standard. Criteria for determining the color component of a digital image providing less disturbance in the quantity of sequential triads in a matrix of unique colors and the basic steps for determining for a particular image the most and least suitable for steganographic transformation color components in terms of changes in color triplets in a matrix of unique colors are given. [ABSTRACT FROM AUTHOR]

The paper proposes a modification of the steganographic method of embedding a digital watermark into Wavelet transform domain of a digital image. The proposed method embeds a watermark in the high-frequency detail coefficients of a three-level Wavelet transform using the Dobeshi-43 wavelet, which ensures the high quality of the resulting stego. Stego is a result of embedding of digital watermark into container and is saved in a losses format that is caused by wide dissemination of such images due to their small size. Discrete cosine transform is applied to a digital watermark, thereby increasing the efficiency of detecting the presence of a watermark in a stego. In work the analysis of efficiency of detecting of the digital watermark depending on the threshold values defining the choice of the detailing coefficients used for embedding is carried out. A result of this analysis is establishment the fact that increase in thresholds without application of Discrete cosine transformation to the watermark leads to considerable deterioration in accuracy of detecting. Conversely, the proposed method with the threshold values Т1=80 and Т2=90 of the choice of detailing coefficients in conjunction with the Discrete Cosine Transformation of the watermark matrix increased the percentage of correctly identified stegos. The results of the experiments showed the resistance of embedding a digital watermark to compression attacks (even with image quality QF=5), imposing noise, sharpening, shifting, for which the detection accuracy averages 97%. Also, the method is resistant to scaling provided QF>50. This paper compares the resistance of the original method of embedding a digital watermark and different variations of the proposed modification of the steganographic method to complex attacks (compression in combination with other types of digital image processing). [ABSTRACT FROM AUTHOR]