A Semi-Blind Watermarking Scheme for Images using a Tree Structure

In this paper, we embed the watermark in a tree structure in the Discrete Wavelet Transform domain. For watermark embedding, the two level DWT decomposition of an NxN gray scale image I is computed. The same PRN sequence is embedded into the DWT coefficients higher than a given threshold T₁ in the LL2 and HH2 bands. The watermark is also embedded into the children of DWT coefficients. The original DWT coefficients are replaced by the modified DWT coefficients. The final step is to compute the inverse DWT to obtain the watermarked image I. For watermark detection, the DWT of the watermarked and possibly attacked image I* is computed. All the DWT coefficients higher than a given threshold T₂ in the LL2 and HH2 bands are selected. Then the sum Z of all attacked DWT coefficients multiplied by either the embedded watermark or other random PRN sequence is computed, divided by the length of the PRN sequence. The sum is also computed for the children of modified DWT coefficients. A predefined threshold T is chosen for LL2 and HH2 bands and the HH1 band. In each band, if Z exceeds T, the conclusion is that the watermark is present.