A New High Robust Digital Image Watermarking Algorithm Using Discrete Wavelet Transform and Recurrent Neural Network

In this paper, a robust watermarking algorithm given recurrent neural networks (RNN) is proposed. We present discrete wavelet transform (DWT) innovation for watermarking. The network is laid out in the DWT domain given the coefficient relations of DWT, RNN is a regularization algorithm that works in light of the different activation functions likes sigmoid, tangent hyperbolic in hidden nodes in bound together environment structure to extract robustness features of DWT coefficients, these features are accustomed to embedding the watermark bits in DWT coefficients emphatically, the robustness features have been utilized to decide the most noteworthy changeable threshold values of the DWT coefficients, interpreted as Possible Probability Criterion (PPC). The coordination of watermarked DWT coefficients save as a key matrix to finding watermark bits in the extraction process. This key expands the security level of the watermarking algorithm. On account of the learning and adaptive abilities of RNN, the trained neural networks can precisely recover the watermark from the watermarked image. A few sorts of attacks are taken to the attacked watermarked image.