Hidden Data Embedding Method Based on the Image Projections Onto the Eigenvectors of Subinterval Matrices

Abstract

We consider the new method of hidden data embedding based on the transform of the container-image using the apparatus of subinterval matrices of the cosine transform. The developed method deals with the analysis of the container-image projections onto the eigenvectors of subinterval matrices. A decisive rule for the choice of informative and non-informative image projections subsets based on a given significance level is proposed. The computational experiments results of projections partitioning into informative and non-informative subsets show that it is possible to obtain a different numbers of informative and non-informative projections subsets using different significance levels. It allows to implement a hidden embedding of different data amounts. The embedding data are represented by a binary sequence. In our method we proposed to implement the data embedding on the basis of a relative change of given projections values. To test the workability of the developed method computational experiments were carried out. Their results showed that the developed method allows to perform data recovery without distortion, and causes a slight distortions of the image containing the embedded data. Also we carried out comparative computational experiments to compare the results of the developed method application with the results of Е. Koch and J. Zhao method and spread spectrum method. Their results showed that the developed method causes less distortions of the container-image than other ones.