Page 25 - 2023-Vol19-Issue2
P. 25
21 | Taheri
Fig. 4. Watermark Embedding Process
the hopfield neural network of associative memory process, Fig. 5. Watermark Extracting Process
the final extracted watermark is gotten through transformed
and restructures according to inverse process. The watermark watermarked image, and results have been integrated in Tables
embedding procedure is shown in Fig. 5. (I-III). In equation (2) OW is the original watermark and EW
is the Extracted Watermark image from separating watermark
III. IMPLEMENTATIONS AND RESULT calculation. Dot operation in this equation is explanatory sum
of product of respective entries. The highlighted rows in each
The original and watermarked images have been displayed Tables show our better algorithm robustness versus previous
in Figs.6-11. Barbara, Baboon, and Cameraman images have methods. The reason of comparison of our algorithm with
been utilized to execute the watermarking calculation. Orig- our previous methods is, these methods have better adaptation
inal watermark is a twofold image and its size is 64 × 64. and robustness versus other methods from other researchers.
The original watermark image is displayed in Fig.12 The
performed attacks on the watermarked images are as per the IV. CONCLUSION
following: Gaussian noise; median filtering 3*3; low pass
filtering; cropping 25% and resizing 1/5 the image; jpeg com- Proposed method achieved PSNR more than mentioned
pression with quality elements of 10, 25, and 75, lastly jpeg methods of DWT-DCT. PSNR signify robustness against vi-
2000 compression with bit rate 3. sual and statistical attacks for invisible watermarking. So IWT
based watermarking process is more imperceptible and robust
SIM(OW, EW ) = OW.EW × 100 (2) for visual attacks. In the proposed work, watermark is em-
OW 2 bedded in high frequency band or edge and noise information
not in characteristics and shapes information of cover image.
PSN R = 10log( (OI(i, 255 W I(i, j)) )2 (3) Further proposed technique may be verified for robustness of
j) - statistical attacks in future. PSNR of recovered watermarked
?i j image is higher compared to existing DWT-DCT technique
so loss of secret information is also reduced using proposed
The gauge of similitude (SIM) or mean square estimate, as technique due to avoid of fraction loss in Integer wavelet trans-
mentioned in abstract, between the extracted watermark image form. The watermark signal is preprocessed by scrambling
and the original watermark image as equation (2), along the technology. Integer wavelet transform plays most roles in the
peak sign to noise ratio (PSNR) of watermarked image (WI), robustness of algorithm, because the integer values of coeffi-
and original image (OI) image will be determined having cients have more margins tolerate about watermarking attacks
played out every last one of the referenced attacks on the
