An embedding approach using orthogonal matrices of the singular value decomposition for image steganography

摘要

This paper aims to reduce the embedding errors, maintain the image fidelity, and reduce the errors, when detecting the embedded messages in images. An embedding approach is proposed that depends on using the orthogonal matrices of the Singular Value Decomposition (SVD) as a vessel for embedding information instead of embedding in the singular values of the images. Three ways are suggested to reduce the embedding errors and maintain the image fidelity, when detecting the embedded message. These ways are increasing the number of columns protected without embedding, choosing the suitable block size to embed in and adjusting the singular values in order to give a high qualify of the stego image. Results show that utilization of the orthogonal matrices of the SVD for information hiding can be as effective as using transform-based techniques, and it gives better results than those obtained with the Least Significant Bit (LSB) technique.