Choice of location for the watermark bits in the cover


The choice of the location of the watermark is crucial for both security and psychovisual impact of the watermarked image. I explore some of the methods related to the choie of location.

Patchwork Algorithm

Suggested by Bender et al.(1995) this asks the binary question “does the person know the key used to build watermark” The insertion process is done as follows: the sender generates using a PRNG n random locations in the image (a(i), b(i) for i = 1 to n (where n is the length of the watermark). Now a(i) = a(i) + 1 ; b(i) = b(i) - 1. The extraction process is to use the secret key on hand to get the random locations and then sum the values at those locations Iff the secret key were the right one the sum would approximately be equal to 2*n else it would be approximately equaly to zero. This is due to the statistical property:

E(s) = sum(E[a(i)] - E[b(i)]) = 0 (for i = 1..n)


Public-key crypto and public-key watermarking

The big drawback with most watermarking methods is the shared-key concept. Ideally the public key crypto concept can be applied to watermarking wherein there are two keys : one of them the private key can be used to watermark the image and the public key to verify it.

Hartung and Girod[18] show how to use public-key techniques using direct seqeunce spread-spectrum models. The direct sequence technique usually requires the spread sequence S for both spreading and unspreading; however due to the robustness of the encoding it is possible to reconstruct the original key without the knowledge of the entire sequence.

Predictive coding for psychovisual watermarking


These are widely used in source coding to predict new value from former value. The assumption is that neighbouring pixels are highly correlated. In the watermarking context predictive coding is useful for its pschovisual effectivness. The human vision system is less accurate in textured/edge regions. However its very sensitive to smooth regions with uniform values; in predictive coding the error conditions matche these properties and the error signal can be used as a modulation carrier for the watermark. It also is in the literature[19] that predictive coding does not need the original image