Spectral Characteristics Analysis of Images Matrix Masking Results

The article describes the results of a computational experiment to assess the capabilities of extracting useful information if an image masked by quasi-orthogonal matrices sent over an open channel became available to a third party. Hadamard and Mersenne matrices of symmetric and cyclic structure are considered. The results confirm the data that images masked by small-sized matrix leaves edges of the original image on the masked image. However, with an increase in the size of the masking matrix, all considered in the article matrices reliably hides the original image during visual analysis. Masking by symmetric Mersenne-Walsh matrices and cyclic Mersenne matrices based on modified M-sequences provides better spectral secrecy of masked images in comparison with Hadamard matrices. Mersenne matrices of cyclic structure, with equal sizes of the image and the masking matrix, bring the phase spectrum of the masked image to a form close in spectrum to uniform noise, which makes their use more preferable based on the considerations that the human visual system is extremely sensitive to phase-frequency distortions of the visual information.

quasi-orthogonal matrices, images matrix masking, Hadamard matrices, Mersenne matrices