With the advent of technology and multimedia production, the world has witnessed a tremendous increase in digital media attacks, which duplicates, forges and tamper the data leading to the violation of copyright laws. In this paper, a robust and secure digital image watermarking is proposed, which exploits the chaotic behaviour of the nonlinear oscillators realized through Memristive diodes. The proposed scheme relies on a Human Visual System (HVS) model in order to mimic the real-life scenario. To improve the robustness of the proposed approach and to further increase the security of the digital watermarked media whilst still retaining compatibility with the real-time events, Histogram of Oriented Gradients (HOG) and extreme learning machine (ELM) is implemented. Secure key generation by means of scrambling through Arnold Transform and the coefficients of Memristive Chaotic Oscillator ensures extreme security. The watermark embedding followed the pixel transformation based on discrete cosine coefficient modification, and a semiblind watermarking extraction procedure was carried out through trained ELM models. A detailed analysis has been presented to evaluate the tradeoff between imperceptibility, security and robustness using performance metrics like PSNR, NC, SSIM, and BER. To establish a real-time implementation of the proposed architecture, the simulated results were verified using real-time chaotic signals generated from the chaotic oscillator, which dictates excellent performance against watermarking attacks and image processing tasks.

No abstract

No abstract