Integrating the watermark bits and the cover bits
The integration step of the watermark and the cover allows a visibility-robustness tradeoff to
be managed in relation to the resepective energies of the watermark and the image
Phase Modulation
Its been shown that phase is robust against noise and contributes significantly to the
psychovisual impact of the image than amplitude. It makes sense hence to use DFT to perform
a phase modulation for robust watermarking. This was proposed first by O'Ruanaidh et al[23]
It must be noted that the DFT in the phase modulation watermarking must respect the
negative symmetry of the DFT transformed image.
Merging that preserves luminance average
This method is based on the classification of image areas by pixel clustering into a set
of homogeneous parts. The algorithm[24] basically divides the image into blocks and then
groups the pixel in each block into regions R1 and R2 namely low/mid and high contrast
regions. These regions are furthere subdivided into two and manipulated so as to preserve
the luminance average while encoding data based on 0 or 1 in the message bit.
Merging Based on DCT coefficient quantization
We have seen how useful DCT is in watermarking. The method used in watermarking is
based on the one described in the Steganography section (Zhao & Koch scheme) of modulating
relative values of two coefficients in an image block.
Merging Based on block substitution in Fractal Coding
Fractal compression is achieved by selecting "similar" domain blocks for
set of range blocks. This is morefully described in [25]. In this approach
the domain and range blocks are partitioned at different resolutions (2Bx2B , BxB typically)
and for each range block the encoder searches for the best collage match from a
suitably transformed and selected domain block. For this search the candidate domain blocks
are usually transformed by performing subsampling, isometry, scale and shift.
For watermarking, the key idea is to restrict the search window. For instance, instead of scanning
the whole image two parts can be defined and according to bit value of the message to be hidden
only one of them would be considered.