Fingerprinting system for still images based on the use of a holographic transform domain

We consider the watermarking method based on a holographic transform domain image proposed by A. Bruckstein. Our testing showed that it is resistant not against all possible attacks declared by his inventor, under the condition of a very high image quality just after WM embedding. Only a small part among 120 bits embedding into the image has an acceptable error probability after extraction if some attacks hold. Therefore we propose to modify this system for fingerprinting where only fixed bits are embedded into the most reliable places of the frequency mask. Systematic linear binary codes with large minimal code distance are used in order to correct errors. Simulation showed that such system provides sufficiently reliable tracing “traitors” under the most types of attacks subjected to remove WM, while keeping a good quality of the image just after embedding.