Trellis-coding embedder and Viterbi detector

The following is an 8-bit blind embedding technique using Trellis codes for image watermarking. The Viterbi algorithm is used to extract the 8 bits of the message.

Embedding

This method is based on System 5 presented in [ 1 ]. It is a similar based method to E-Simple-8, in the sense that 8 different marks are embedded in order to hide 8 bits. However, in this case Trellis codes are used to encode the bits, which allows for a much more reliable decoding.

It is a robust method, i.e. able to withstand some image alterations.

The code for the embedding method is available at the following link: E_TRELLIS_8.

To embed a mark, just run a command like the following:

$ python3 E_TRELLIS_8.py image.png passw123 10101010 marked.png

Detection

Mark detection and decoding is performed by calculating the linear correlation between the marked image and each of the eight marks.

The formula used for the linear correlation is as follows:

$Z_{lc} = \frac{1}{N} \bar{c} · \bar{w_r} = \frac{1}{N}\sum_{xy}c(x,y) w_r(x,y)$

The Viterbi algorithm is used for message decoding.

The code of the detection method is available at the following link: D_TRELLIS_8.

To detect the mark and see if a 0 or a 1 has been embedded, just run a command like the following:

$ python3 D_TRELLIS_8.py marked.png passw123
msg: 10101010

References

1. I. J. Cox, M. L. Miller, J. A. Bloom, J. Fridrich y T. Kalker (2008). Digital Watermarking and Steganography. Morgan Kaufmann. Second Edition.