Robust Watermarking Method for Colour Still Image Databases
Mihai Mitrea, Francoise J. Preteux, Adriana Vlad · 2002
This paper addresses the issues of oblivious, transparent and Stirmark robust watermarking of colour still image databases. Note that the well-known Cox and Ó Ruanaidh non-oblivious methods are devoted only to grey levels individual still images and, unfortunately, they do not survive the Stirmark attack. Besides, the Cox method allows only a single bit to be recovered at the detection side (a marked/unmarked answer is obtained). By extending and generalising the principles of the Cox and Ó Ruanaidh approaches, and by overcoming their limitations, the new oblivious method proposed in this paper allows a 64 bit message to be checked out at the detector. The message is coded according to the Spread Spectrum – Code Division Multiple Access (SS–CDMA) technique, and embedded into the most significant DCT (Discrete Cosine Transform) coefficients (except for the DC – the direct component). Applied to 1000 sequences of 32 still images, the method exhibits an excellent behaviour in terms of transparency (both fidelity and quality), robustness (JPEG compression, the change of format, scaling, print&scan, Stirmark attack) and probability of false alarm (lower than 10102 − ×). Finally, the extension towards colour video watermarking is taken into consideration; some solutions are proposed and experimental results are presented. 1.