I am implementing wet-paper codes (WPC) with randomly generated parity-check matrix $H$, based on this paper. As the wet DCT coefficients, I set DCT coefficients with value 0, or with values 0 and 1 (trying both).

For almost every cover image, the input vector at iteration contains too many wet pixels and the program fails to find the vector v. What is the way to deal with that? Do I even choose the wet DCT coefficients correctly?

  • $\begingroup$ that reference is paywalled. you need to expand on your question $\endgroup$
    – kodlu
    Commented Oct 14, 2022 at 12:13
  • $\begingroup$ Huh? I could download without issue, though I do have research gate access. Maybe that matters or not. $\endgroup$
    – Maarten Bodewes
    Commented Oct 21, 2022 at 15:24
  • $\begingroup$ I fixed the link, no payment/access is needed now. $\endgroup$ Commented Oct 22, 2022 at 8:02

1 Answer 1


It is not possible to hide a message, without modifying the wet pixels, if there are too many wet pixels. Think for example, of an image with all pixels wet: you can't modify the image without modifying some wet pixel.

The only options in these cases are to modify the wet pixels or hide less information.

Other more advanced options, such as STC, allow you to indicate a cost for each pixel. Therefore, the modification of the pixels with the lowest cost will be prioritized. In this case, the equivalent of the wet pixel is the pixel with infinite cost. But even in this case, these pixels can end up being changed if there are too many of them or if the message is too big.

You can learn more about this in the following paper, in which the amount of modified wet pixels is studied.

Minimizing Additive Distortion in Steganography Using Syndrome-Trellis Codes by T. Filler, J. Judas and J. Fridrich. [ PDF ].


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.