Authentication of DICOM medical images using independent component analysis (ICA)

摘要

This paper proposes a blind content-based watermarking scheme for image authentication using ICA and DCT. The watermark to be embedded is obtained from the host image itself in terms of the Frobenius norm of the mixing matrix obtained during ICA. These are embedded in the mid-frequency DCT coefficients. The proposed method correctly authenticates the image even under normal image processing operations and it correctly detects tampering and identifies the tampered regions of the image. The average PSNR value in this proposed ICA-based technique is around 97 which makes it superior than the existing watermarking techniques.