Improvement of Robustness and Perceptual Quality of Image Watermarking using Multi-Objective Evolutionary Optimizer

This paper proposes optimized Discrete Wavelet Transform (DWT) and Singular Value Decomposition (SVD) watermarking technique using multi-objective evolutionary algorithm using Fibonacci-Lucas transform. In any image watermarking technique, there is always challenge for researcher to achieve perceptual quality and robustness simultaneously under high payload scenario because these two quality metrics are inverse of each other. We achieved optimized ‘Peak Signal to Noise Ratio’ (PSNR) and ‘Normalized Correlation’ (NC) in DWT-SVD domain applying seven different security levels including FibonacciLucas transformation based watermark scrambling. The selection of wavelet for decomposition of cover image is done after testing practical performance various wavelets. The technique is non-blind and tested with 512x512 size cover images and 256x256 size grey scale watermark. We could achieve normalized correlation equals to 1 for all cover images indicating exact recovery of watermark. We got PSNR 79.8611 for Lena, 87.8446 for peppers and 93.2853 for lake images when scale factor K was varied between 1 to 5. The technique undergone 17 various attacks and achieved extracted watermark with normalized correlation more than 0.99 for majority of attacks. The significant improvement in performance metrics is found with compared to existing DWT-SVD based image watermarking techniques in the literature under consideration.