The SCFT representation of color images, introduced by McCabe et al.
In this paper, we propose the use of a frequency-bandpass SCFT domain to conceal the watermark payload in this band.
Once color-based processing is carried out on [A*(u, v) + j B*(u, v)], a* and b* can be recovered using the inverse SCFT transform :
The proposed color image watermarking scheme uses a bandpass representation of the SCFT transform in conjunction with direct-sequence spread-spectrum watermark embedding to mitigate the effects of low signal-to-noise ratio (SNR) on the watermark bits.
Apply forward SCFT transform on the complex-valued quantity, [a*(x, y) +j b *(x, y)], using Equation (4) to get the SCFT quantities [A *(u, v) +jB*(u, v)].
i] are the ith samples of the host and watermarked SCFT quantities [A *(u, v) + jB*(u, v) and [[(A*(u, v)).
To ensure watermark imperceptibility, only a band of the SCFT coefficients, defined in Equation (4), are used for watermark embedding.
In this paper, we model the watermarked SCFT coefficients using the Weibull distribution to blindly recover the embedded bits.
Scalar Watermarking Game: All SCFT host coefficients belong to a single Gaussian channel where all coefficients have equal local variance set to [[sigma].
Parallel Watermarking Game: In this configuration, the SCFT host coefficients are assigned to different (parallel) Gaussian channels based on the level of their local variances.
To allow energy-aware watermark embedding, the SCFT coefficients are classified into L parallel Gaussian channels based on their local variances [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].
3], a spike model is used where the SCFT coefficients are classified into two separate channels using a coarse quantization with threshold equal to 2[D.