Optimized SVD-based robust watermarking in the fractional Fourier domain

Abstract

Digital watermarking is one of the most effective methods for protecting multimedia from different kind of threats. It has been used for many purposes, like copyright protection, ownership identification, tamper detection, etc. Many watermarking applications require embedding techniques that provide robustness against common watermarking attacks, like compression, noise, filtering, etc. In this paper, an optimized robust watermarking method is proposed using Fractional Fourier Transform and Singular Value Decomposition. The approach provides a secure way for watermarking through the embedding parameters that are required for the watermark extraction. It is a block-based method, where each watermark bit is embedded in its corresponding image block. First, the transform is applied to each block, and then the singular values are evaluated through which the embedding modification is performed. The optimum fractional powers, of the transform, and the embedding strength factor are evaluated through a Meta-heuristic optimization to optimize the watermark imperceptibility and robustness. The Artificial Bee Colony is used as the Meta-heuristic optimization method. A fitness function is employed, at the optimization process, through which the maximum achievable robustness can be provided without degrading the watermarking quality below a predetermined quality threshold Qth. The effectiveness of the proposed method is demonstrated through a comparison with recent watermarking techniques in terms of the watermarking performance. The watermarking quality and robustness are evaluated for different quality threshold values. Experimental results show that the proposed approach achieves a better quality compared to that of other existing watermarking methods. On the other hand, the robustness is examined against the most common applied attacks. It is noticed that the proposed method can achieve a higher robustness degree when decreasing the quality threshold value.