УСОВЕРШЕНСТВОВАНИЕ СТЕГАНОГРАФИЧЕСКОГО АЛГОРИТМА, ОСНОВАННОГО НА SIGN-НЕЧУВСТВИТЕЛЬНОСТИ СИНГУЛЯРНЫХ ВЕКТОРОВ БЛОКОВ МАТРИЦЫ ИЗОБРАЖЕНИЯ

In this paper the new method for digital image processing is proposed. Application of this method leads to decrease the conditioning number of the majority of matrix blocks obtained by standard partitioning as a result of preconditioning of the corresponding matrices. The proposed method preserves the reliability of the perception of the distorted image, which in the work is quantitatively estimated using the peak signal-to-noise ratio. The application of the proposed method as pre-processing of the cover image before steganography transformation makes it possible to extend the application domain of the steganography algorithm based on the sign-insensitivity of the singular vectors of blocks of the digital image matrix developed by one of the authors earlier. The preconditioning of the block matrix occurred after its precoding. The result of precoding is a symmetric positive-definite sparse matrix, which is mapped to the image block. The positive definiteness and sparsification of obtained matrix make it possible to use as its preconditioner the incomplete Cholesky factorization. The results of the computational experiment are presented. [ABSTRACT FROM AUTHOR]