How do I develop MATLAB code for adding a cryptowatermark in medical images?

I am doing a project on cryptowatermark in medical images.

Please help to give some ideas to develop the MATLAB code.

[Doug Hull]

Please be more specific. No one will know how to answer this. Also, what is crypto watermarking? Likely people can help with the more general aspects of it if it is defined.

[chanpreet]

could anyone help with rsa algorithm?

[Walter Roberson]

Sorry, no, discussion of strong encryption is not allowed on this site, for legal reasons.

Answer 1

[Jan]

The medical images I'm working with are not allowed to be modified in any way. Even the lossless JPEG compression is forbidden, because the decompressed image is not bit-identical to the original due to rounding errors.

Therefore I think of a HMAC-sign, when I hear "crypto" together with "medical image". But this is not a watermarking method.

[Walter Roberson]

I did read a paper in which they did a study of whether lossy JPG was acceptable to the radiologists for some X-Ray images. They did blind tests to eliminate the "I know it is worse!" bias. The study did find that using too much loss was detectable, but found that there was a level of compression below which even the strictest of their radiologists was unable to determine any difference between original and reconstructed image when they were available side-by-side. I do not recall the compression limit determined by the study, but it was enough to be "useful".

Yes, there are fears that "maybe" something would get missed, but the same fears apply to digital images instead of analog. The same fears are routinely dealt with in medical practice with respect to not always using the test or procedure with the highest level of accuracy or specificity (when that test might take many times longer and be many times more expensive); the same kind of fears are dealt with with respect to why one does not always use "the best surgeon in the country".

The organization I work for develops new diagnostic techniques, and it is not uncommon for us to find something that is "better" than the studied statistics for whatever we are dealing with. Much of what we develop only makes it as far as a journal paper, because the improvements might not be significant enough to be marketing. It takes about $2 million to bring a new medical product to market, and our method might require $1 million in equipment (and half a million dollars in building renovations to install the equipment in a proper facility), so if our improvement is (say) 0.3%, it just isn't worth it.

The medical industry and the courts have never demanded that "best possible" be used at each step.

Thus, if an "image watermark" could be developed that could be demonstrated (through studies or through information theory) to have no affect on the diagnostic power of the medical image, then it would be permitted, at least in some countries.