Exploiting De-Noising Convolutional Neural Networks DnCNNs for an Efficient Watermarking Scheme: A Case for Information Retrieval

Digital watermarking has various applications in which it plays an important role; these include document security and multimedia vendor copyrights. This paper presents an efficient watermarking technique called 'modified selective embedding in low frequency' (M-SELF) implementing a Fast Fourier Transform (FFT) for frequency domain analysis where FFT shift brings down lower-frequency components to the center of the cover image. Embedding of the watermark is done by optimal radial implementation after coordinates computation via phase and magnitude separation. Furthermore, the watermark is protected by the convolutional code (1/2) and Advanced Encryption Standard (AES-128) algorithm providing a double layer of security. Moreover, the exploitation of de-noising convolutional neural networks (DnCNNs) is another contribution of the proposed work. The network is utilized as a de-noiser in an integrated manner for the proposed watermarking technique on blind level of additive white Gaussian noise (AWGN) to check the robustness of the technique. Performance evaluation of the proposed integrated watermarking technique is done in terms of percentage of bits being retrieved as a success rate at each level of noise and perceived visual quality by using full-reference image quality assessment (IQA) metrics such as peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM).