Digital Watermarking

                                  by
                             Arak Sutivong

Digital watermarking has recently emerged as a candidate solution to
the difficult problem of providing guarantees to creators and
consumers of digital content about the protection of their legal
rights.  Although the variety of proposed algorithms is large, most
schemes are derived from spread spectrum methods.  Such schemes have
been shown to be robust to various attacks [1].  The goal of this
project is to embed a reasonably large amount of watermark information
(1000-2000 bits) into a 512x512 image using a spread spectrum method.
The originality of this work lies in the way information is embedded
into the host image.  Specifically, instead of using a random
signature to represent a watermark as done in [1], the proposed method
assigns a unique signature (i.e., Walsh function) to each bit of the
watermark message.  The modulated waveforms (one for each watermark
bit) are then combined together.  For security purposes, the watermark
is then scrambled (i.e., masked by another random sequence) prior to
being added to the host signal.  This process is similar to the way
information from different users is transmitted on the forward link of
the IS-95 CDMA system.  Because of the orthogonality of Walsh
functions, each information bit can be easily extracted at the
receiver (provided that a copy of the unwatermarked image is also
available).  The watermark embedding is done in a DCT domain to
increase the robustness against various attacks.

If time permits, I'm hoping to do a detection without prior knowledge
of the unwatermarked image.  This is a substantially harder
problem. The information watermark carries will probably be
substantially less for the same kind of detection reliability.  The
problem arises primarily because of the unknown bias introduced by the
host signal (during the detection process), which results in an
ambiguous decision boundary.


[1] I. Cox, J. Kilian, T. Leighton, and T. Shamoon, "Secure Spread
Spectrum Watermarking for Multimedia," IEEE Transaction on Image
Processing, vol. 6, no. 12, pp 1673-1687, 1997.